Skip to content

Do the "preprocessing" right for PyTorch compiled grouped GEMM #513

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Open
wants to merge 1 commit into
base: main
Choose a base branch
from
Open

Do the "preprocessing" right for PyTorch compiled grouped GEMM #513

wants to merge 1 commit into from

Conversation

alexsamardzic
Copy link

No description provided.

@meta-cla meta-cla bot added the cla signed label Oct 3, 2025
@alexsamardzic
Copy link
Author

Currently, torch_compile_grouped_gemm and preprocessed_pt2_triton_grouped_mm are the same.

I think there is no point to benchmark torch_compile_grouped_gemm, as it takes into account both the auto-tuning and "preprocessing" arguments. On the other hand, with the change in this PR, preprocessed_pt2_triton_grouped_mm is on par with preprocessed_aten_grouped_mm, which is expected.

(I believe the point about auto-tuning holds for triton_grouped_gemm too.)

@xuzhao9 xuzhao9 requested a review from NikhilAPatel October 3, 2025 17:48
@xuzhao9
Copy link
Contributor

xuzhao9 commented Oct 3, 2025

cc @NikhilAPatel, can you help take a look?

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

@NikhilAPatel NikhilAPatel Awaiting requested review from NikhilAPatel

At least 1 approving review is required to merge this pull request.

Assignees
No one assigned
Labels
cla signed
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
2 participants
@alexsamardzic @xuzhao9