Skip to content

Conversation

@garroud
Copy link
Contributor

@garroud garroud commented Nov 18, 2025

Summary:
X-link: https://github.com/facebookresearch/FBGEMM/pull/2146

att. When all the input embedding are from the same device, we can just use cat as a short cut. This can avoid unnecessary cross device sync with current impl.

Differential Revision: D87306514

@meta-cla meta-cla bot added the cla signed label Nov 18, 2025
@meta-codesync
Copy link
Contributor

meta-codesync bot commented Nov 18, 2025

@garroud has exported this pull request. If you are a Meta employee, you can view the originating Diff in D87306514.

Summary:

X-link: facebookresearch/FBGEMM#2146

att. When all the input embedding are from the same device, we can just use cat as a short cut. This can avoid unnecessary cross device sync with current impl.

Differential Revision: D87306514
@meta-codesync
Copy link
Contributor

meta-codesync bot commented Nov 20, 2025

This pull request has been merged in 7826ec9.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants