[bugfix] Fix page size for create_flashmla_kv_indices_triton() for cutlass mla #8685
Conversation
There was a problem hiding this comment.
Summary of Changes
Hello @trevor-m, I'm Gemini Code Assist1! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed!
This pull request resolves a critical bug in the cutlass_mla attention backend that was causing a severe accuracy degradation (to 0%) in the system. The root cause was an incorrect passing of the PAGE_SIZE argument to a core function, triggered by a recent change in the function's argument order. The fix ensures the correct parameter is always used by explicitly naming it, making the code more resilient to future interface changes.
Highlights
- Bug Fix for Accuracy: Addressed a critical bug in the
cutlass_mlabackend where thePAGE_SIZEargument was being incorrectly passed to thecreate_flashmla_kv_indices_tritonfunction. This issue, caused by a recent change in argument order, resulted in a severe accuracy degradation (0%). - Argument Passing Robustness: Improved the robustness of argument passing by explicitly using a named argument (
PAGED_SIZE=PAGE_SIZE) when callingcreate_flashmla_kv_indices_triton. This prevents future regressions if the function's argument order changes again.
Using Gemini Code Assist
The full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips.
Invoking Gemini
You can request assistance from Gemini at any point in your pull request via creating an issue comment (i.e. comment on the pull request page) using either /gemini <command> or @gemini-code-assist <command>. Below is a summary of the supported commands.
| Feature | Command | Description |
|---|---|---|
| Code Review | /gemini review |
Performs a code review for the current pull request in its current state. |
| Pull Request Summary | /gemini summary |
Provides a summary of the current pull request in its current state. |
| Comment | @gemini-code-assist | Responds in comments when explicitly tagged, both in issue comments and review comments. |
| Help | /gemini help |
Displays a list of available commands. |
Customization
To customize Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a .gemini/ folder in the base of the repository. Detailed instructions can be found here.
Limitations & Feedback
Gemini Code Assist may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with 👍 and 👎 on @gemini-code-assist comments or fill out our survey to provide feedback.
You can also get AI-powered code generation, chat, as well as code reviews directly in the IDE at no cost with the Gemini Code Assist IDE Extension.
Footnotes
-
Review the Privacy Notices, Generative AI Prohibited Use Policy, Terms of Service, and learn how to configure Gemini Code Assist in GitHub here. Gemini can make mistakes, so double check it and use code with caution. ↩
Motivation
#8632 added a new arg to
create_flashinfer_kv_indices_tritonwhich affected the arg order, so page size was no longer passed correctly to the function in the cutlass mla backend. This caused the accuracy to go to 0.Modifications
Use named arg for page size so the order change won't affect it.
Accuracy Test
BEFORE
AFTER (note: #8364 will fix a separate issue and raise this to 0.965)
Benchmark & Profiling
No perf impact.
Checklist