Skip to content

Disable Flash Attention GQA support on ROCM#133884

Closed
xinyazhang wants to merge 2 commits intopytorch:mainfrom
ROCm:xinyazhang/nogqa-2.5main
Closed

Disable Flash Attention GQA support on ROCM#133884
xinyazhang wants to merge 2 commits intopytorch:mainfrom
ROCm:xinyazhang/nogqa-2.5main

Conversation

@xinyazhang
Copy link
Collaborator

@xinyazhang xinyazhang commented Aug 19, 2024

Currently GQA is unsupported on ROCM.

Partially addresses #133540

cc @jeffdaily @sunway513 @jithunnair-amd @pruthvistony @ROCmSupport @dllehr-amd @jataylo @hongxiayang

@pytorch-bot
Copy link

pytorch-bot bot commented Aug 19, 2024

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/133884

Note: Links to docs will display an error until the docs builds have been completed.

❌ 1 New Failure, 1 Unrelated Failure

As of commit 3246e80 with merge base f31404b (image):

NEW FAILURE - The following job has failed:

FLAKY - The following job failed but was likely due to flakiness present on trunk:

This comment was automatically generated by Dr. CI and updates every 15 minutes.

@pytorch-bot pytorch-bot bot added the module: rocm AMD GPU support for Pytorch label Aug 19, 2024
@pruthvistony pruthvistony added ciflow/rocm Trigger "default" config CI on ROCm rocm This tag is for PRs from ROCm team rocm priority high priority ROCm PRs from performance or other aspects labels Aug 19, 2024
@jithunnair-amd
Copy link
Collaborator

Closing this in favor of #134498, which will include the GQA-related changes.

@xinyazhang
Copy link
Collaborator Author

Already part of AOTriton 0.7b integration PR

@xinyazhang xinyazhang closed this Sep 12, 2024
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

ciflow/rocm Trigger "default" config CI on ROCm module: rocm AMD GPU support for Pytorch open source rocm priority high priority ROCm PRs from performance or other aspects rocm This tag is for PRs from ROCm team

Projects

None yet

Development

Successfully merging this pull request may close these issues.

4 participants