Skip to content

Conversation

@bursteratom
Copy link
Contributor

@bursteratom bursteratom commented Apr 7, 2025

Bumping flex attention monkeypatch to transformers 4.51

@bursteratom bursteratom requested a review from winglian April 7, 2025 19:15
@SalmanMohammadi
Copy link
Contributor

huggingface/transformers#37285 has landed btw

@bursteratom bursteratom force-pushed the flex_patching_update branch from b98dbaf to cdb1606 Compare April 8, 2025 13:23
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I don't know that this can be removed yet, as the released version in transformers had an edge case that affects most of our docker builds. huggingface/transformers#37399

Comment on lines -909 to -913
if self.cfg.flex_attention:
self.model_kwargs["attn_implementation"] = "flex_attention"
self.model_config._attn_implementation = ( # pylint: disable=protected-access
"flex_attention"
)
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Do we need to at least keep these lines?

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Do we still even need this PR?

@winglian
Copy link
Collaborator

superseded by #2469

@winglian winglian closed this Apr 28, 2025
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

5 participants