Skip to content

Skip Tokamax RaggedDotGroupSizes for FP8#3420

Merged
copybara-service[bot] merged 1 commit intomainfrom
amandaliang
Mar 16, 2026
Merged

Skip Tokamax RaggedDotGroupSizes for FP8#3420
copybara-service[bot] merged 1 commit intomainfrom
amandaliang

Conversation

@BirdsOfAFthr
Copy link
Copy Markdown
Collaborator

@BirdsOfAFthr BirdsOfAFthr commented Mar 14, 2026

Description

FP8 path is still using tokamax internal backend APIs. The new RaggedDotGroupSizes was introduced (pull3330) for Tokamax public APIs in bf16 path, which broke FP8.

Tests

Benchmarks were run internally.

Checklist

Before submitting this PR, please make sure (put X in square brackets):

  • I have performed a self-review of my code. For an optional AI review, add the gemini-review label.
  • I have necessary comments in my code, particularly in hard-to-understand areas.
  • I have run end-to-end tests tests and provided workload links above if applicable.
  • I have made or will make corresponding changes to the doc if needed, including adding new documentation pages to the relevant Table of Contents (toctree directive) as explained in our documentation.

FP8 path is still using tokamax internal backend APIs. The new `RaggedDotGroupSizes`` was introduced ([pull3330](#3330)) for Tokamax public APIs in bf16 path, which broke FP8.
@BirdsOfAFthr BirdsOfAFthr changed the title Skip Tokamax RaggedDotGroupSizes for fp8 Skip Tokamax RaggedDotGroupSizes for FP8 Mar 14, 2026
Copy link
Copy Markdown
Collaborator

@shuningjin shuningjin left a comment

Choose a reason for hiding this comment

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

LGTM. This aligns with my experiments.

@codecov
Copy link
Copy Markdown

codecov bot commented Mar 14, 2026

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

@copybara-service copybara-service bot merged commit a3c19fd into main Mar 16, 2026
59 checks passed
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.

4 participants