fix race condition bug in cute _flash_attn_fwd in multiple gpu env #1793
+48
−45
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
I'm not sure it's "race condition", I suspect it just launches the kernel on cuda device 0 even when the data is on cuda device 1 |
|
I see. When do you expect we will have Blackwell support for _flash_attn_bwd? |
|
3-4 weeks |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
In multi-gpu runs, the cute implementation of _flash_attn_fwd returns incorrect values for gpu != 0. This can be fixed with torch.cuda.device context manager as sugguested issue1782: