[Perf] Enable missing FlashInfer MoE autotuning#290
Conversation
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: cadaca8945
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
Signed-off-by: Mohammad Miadh Angkad <176301910+mmangkad@users.noreply.github.com>
mmangkad
force-pushed
the
enable-flashinfer-moe-autotune
branch
from
cadaca8 to
6ccb610
Compare
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 6ccb6103d5
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
1 participant
Summary
I noticed some FlashInfer MoE backends were missing the existing first-call autotune flow. This PR enables it for FP8 Cutlass, NVFP4 Cutlass, and NVFP4 TRTLLM.
Test Plan
Manually validated on GB300; CI is the primary validation.