[BUGFIX] Skip Mamba Cache Slot 0 to Avoid Using Dummy Cache

[Chen-0210]

Motivation

During CUDA Graph capture, dummy mamba cache may be written to slot 0.
self.free_slots in mamba_pool does not exclude slot 0. So slot 0 may contain garbage values at first if cuda graph is enabled.

Previously, this bug was hidden by the server-warm-up — ”The capital city of France is“.

Modifications

Exclude slot 0 from the free slot list to prevent dummy mamba cache being used in prefix cache.
cc @yizhang2077

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.

Review Process

1. Ping Merge Oncalls to start the PR flow. See the PR Merge Process.
2. Get approvals from CODEOWNERS and other reviewers.
3. Trigger CI tests with comments or contact authorized users to do so.
   • /tag-run-ci-label, /rerun-failed-ci, /tag-and-rerun-ci
4. After green CI and required approvals, ask Merge Oncalls to merge.

[gemini-code-assist]

Summary of Changes

Hello @Chen-0210, 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 bug in the Mamba cache management system. During CUDA Graph capture, dummy data could be written to the first cache slot (index 0). To prevent this dummy data from being mistakenly utilized in the prefix cache, the system now ensures that slot 0 is never added to the list of available free slots. This change enhances the reliability of the cache by preventing potential data corruption from padded tokens.

Highlights

• Mamba Cache Bugfix: Addresses an issue where dummy data written to cache slot 0 during CUDA Graph capture could be inadvertently used in the prefix cache, leading to incorrect behavior.
• Cache Slot Exclusion: The initialization of free cache slots has been modified to explicitly exclude slot 0, preventing it from being allocated for actual data storage.

🧠 New Feature in Public Preview: You can now enable Memory to help Gemini Code Assist learn from your team's feedback. This makes future code reviews more consistent and personalized to your project's style. Click here to enable Memory in your admin console.

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 bug in the Mamba cache where dummy data from slot 0 could be used. The fix correctly initializes the free slot list to exclude slot 0. My review points out a related issue in the clear() method which was not updated and would re-introduce the bug. I've suggested a fix to ensure consistency.

[yizhang2077]

once ci passed, it can be merged

[yizhang2077]

/tag-run-ci-label

[Chen-0210]

/rerun-failed-ci

[ispobock]

/tag-and-rerun-ci