Fix hanging during DeepGemm Warmup

[Fridge003]

Motivation

co-authored: @ch-wan
Still working on this PR. The cutoff max_m should be something more flexible

Modifications

Accuracy Tests

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.
• Follow the SGLang code style guidance.
• Work with maintainers to merge your PR. See the PR Merge Process

[gemini-code-assist]

Summary of Changes

Hello @Fridge003, 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 a critical issue where the DeepGemm warmup process could hang, particularly on systems with limited GPU memory. By implementing a dynamic adjustment of the 'max_m' parameter during precompilation, the system now intelligently caps memory usage based on available GPU resources, preventing out-of-memory conditions. This enhancement not only resolves the hanging problem but also allows for the re-introduction of a previously disabled test case, signifying increased robustness in the DeepGemm module.

Highlights

• Dynamic Memory Management for DeepGemm Warmup: The pull request introduces a mechanism to dynamically adjust the 'max_m' parameter during DeepGemm precompilation based on the available GPU memory.
• Prevention of Out-of-Memory (OOM) Hangs: If the available GPU memory is less than 10GB, the 'max_m' value is capped at 4096 to prevent potential out-of-memory errors that could cause the DeepGemm warmup process to hang.
• Test Suite Enhancement: A previously skipped test case, 'TestDeepseekMTP', which was prone to timeouts on CI runners, has been re-enabled, indicating improved stability and correctness of the DeepGemm warmup process.

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 by creating a comment using either /gemini <command> or @gemini-code-assist <command>. Below is a summary of the supported commands on the current page.

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 pull request 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 addresses a hanging issue during DeepGemm warmup by introducing a memory-aware limit on the max_m parameter. When available GPU memory is low, max_m is capped to prevent out-of-memory errors that could cause the process to hang. This is a good approach to improve robustness. The test changes correctly enable the precompilation logic to verify the fix. My main feedback is to replace the magic numbers used for the memory threshold and max_m cap with named constants to improve code clarity and maintainability.