Simplify GQA conditions in sdpa_attention.py#41699
Merged
vasqu merged 1 commit intohuggingface:mainfrom Oct 17, 2025
Merged
Conversation
Removed unnecessary checks for key being a torch.fx.Proxy in GQA conditions because fx tracing is no longer supported, and torch.export supports enable_gqa.
5 tasks
justinchuby
commented
Oct 17, 2025
| @@ -32,13 +32,11 @@ def use_gqa_in_sdpa(attention_mask: Optional[torch.Tensor], key: torch.Tensor) - | |||
| # 1.cuda or Ascend NPU | |||
| # - torch version >= 2.5 | |||
| # - attention_mask is None (otherwise it will fall back to the math kernel) | |||
Contributor
Author
There was a problem hiding this comment.
Is it required that attention_mask is None? Just wanting to check it the situation has changed
Contributor
There was a problem hiding this comment.
Yup, at least on older versions it is! Was only recently relaxed
Contributor
There was a problem hiding this comment.
See the description here https://docs.pytorch.org/docs/2.5/generated/torch.nn.functional.scaled_dot_product_attention.html (under gqa --> supports only flash and math --> flash only works without a mask)
We want to avoid the math kernel
|
The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update. |
ngazagna-qc
pushed a commit
to ngazagna-qc/transformers
that referenced
this pull request
Oct 23, 2025
Removed unnecessary checks for key being a torch.fx.Proxy in GQA conditions because fx tracing is no longer supported, and torch.export supports enable_gqa.
SangbumChoi
pushed a commit
to SangbumChoi/transformers
that referenced
this pull request
Jan 23, 2026
Removed unnecessary checks for key being a torch.fx.Proxy in GQA conditions because fx tracing is no longer supported, and torch.export supports enable_gqa.
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
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.
Removed unnecessary checks for key being a torch.fx.Proxy in GQA conditions because fx tracing is no longer supported, and torch.export supports enable_gqa.