Skip to content

[Optim](gpt-oss): mv padding to fused rmsnorm in tp 1#407

Open
PerryZhang01 wants to merge 2 commits intomainfrom
remove_pad
Open

[Optim](gpt-oss): mv padding to fused rmsnorm in tp 1#407
PerryZhang01 wants to merge 2 commits intomainfrom
remove_pad

Conversation

@PerryZhang01
Copy link
Contributor

@PerryZhang01 PerryZhang01 commented Mar 25, 2026

Motivation

there is not allreduce kernel in tp 1, so rmsnorm can fuse pad kernel, and the input scale is for a8w4 weights

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.

2 participants