Add flash attention as custom op so dynamo can trace it #6875
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
JackCaoG
commented
Apr 2, 2024
| def flash_attention_non_xla(q: torch.Tensor, | ||
| k: torch.Tensor, | ||
| v: torch.Tensor, | ||
| causal: bool = False): |
Collaborator
Author
There was a problem hiding this comment.
I didn't add segment_ids as it is not implemented in flash_attention. Maybe I should add it here... @alanwaketan is flash_attention also a torch.tensor?
Collaborator
There was a problem hiding this comment.
I don't think we need to support segment_ids. We can safely ignore it. I will drop the comment and the parameter later.
Collaborator
Author
|
@alanwaketan can I get a review for this one? |
JackCaoG
added a commit
that referenced
this pull request
Apr 4, 2024
JackCaoG
added a commit
that referenced
this pull request
Apr 5, 2024
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Outdated suggestions cannot be applied.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.
No description provided.