Disable kernel cutlass_mla_decode on SM103

[hlu1]

Motivation

Half of the accuracy tests of cutlass_mla_decode is failing on SM103. Make it throw an exception on SM103 for now until the cause of the failure can be identified and fixed.

Accuracy Tests

Full accuracy tests results are here: https://gist.github.com/hlu1/405371e28ea236150eae4a65c751c8f9

Benchmarking and Profiling

Checklist

• Format your code according to the Format code with pre-commit.
• Add unit tests according to the Run and add unit tests.
• Update documentation according to Write documentations.
• Provide accuracy and speed benchmark results according to Test the accuracy and Benchmark the speed.

[gemini-code-assist]

Summary of Changes

Hello @hlu1, 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 addresses critical accuracy issues encountered with the cutlass_mla_decode kernel on SM103 GPU architectures. To prevent incorrect computations and provide a clear error state, the kernel has been temporarily disabled for SM103 devices, and the corresponding test suite has been adjusted to reflect this change. This ensures stability and prevents the use of a known-faulty kernel on specific hardware until a proper fix can be implemented.

Highlights

• Kernel Disablement: The cutlass_mla_decode kernel is now explicitly disabled for SM103 (compute capability 10.3) devices at runtime due to observed accuracy failures.
• Runtime Validation: A TORCH_CHECK has been added within the cutlass_mla_decode function to ensure it only executes on SM100 devices, preventing its use on SM103 until accuracy issues are resolved.
• Test Suite Adjustment: The test_cutlass_mla.py suite has been updated to skip cutlass_mla_decode tests on SM103 devices, aligning with the kernel's runtime disablement.

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. If you're interested in giving your feedback about your experience with Gemini Code Assist for Github and other Google products, sign up here.

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

1. 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. ↩

[gemini-code-assist]

Code Review

This pull request disables the cutlass_mla_decode kernel on architectures other than SM 10.0 (compute capability 10.0) due to accuracy failures on SM103. The changes correctly add a check in the CUDA kernel and update the corresponding Python tests to skip execution on unsupported hardware. My feedback focuses on improving the error message for better diagnostics when the kernel is run on an unsupported device.