[FlashInfer] Bump FlashInfer version from 0.6.2 to 0.6.3

[mmangkad]

Motivation

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.

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 @mmangkad, 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 focuses on a routine but crucial update: bumping the FlashInfer library to its latest version, 0.6.3. This ensures that the project benefits from any performance improvements, bug fixes, or new features introduced in the updated FlashInfer release. The changes span across dependency declarations, Docker build arguments, and internal version checks, aiming to maintain a robust and up-to-date dependency ecosystem for the project.

Highlights

• FlashInfer Version Update: The FlashInfer library version has been consistently updated from 0.6.2 to 0.6.3 across various configuration and dependency files.
• Dependency Management: The pyproject.toml file now specifies flashinfer_python and flashinfer_cubin at version 0.6.3, ensuring the project uses the latest compatible FlashInfer packages.
• Version Checks and Documentation: Internal version assertion logic and documentation examples for FlashInfer have been updated to reflect the new 0.6.3 requirement, including for features like flashinfer_trtllm.
• CI/CD Alignment: The CI dependency installation script has been modified to fetch and use FlashInfer version 0.6.3, maintaining consistency across development, testing, and deployment environments.

🧠 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.

Changelog

• docker/Dockerfile
  • Updated the FLASHINFER_VERSION ARG from 0.6.2 to 0.6.3.
• python/pyproject.toml
  • Bumped flashinfer_python and flashinfer_cubin dependencies to 0.6.3.
• python/sglang/srt/entrypoints/engine.py
  • Updated the required flashinfer_python version in the assert_pkg_version call to 0.6.3.
• python/sglang/srt/server_args.py
  • Adjusted the minimum flashinfer-python version check for enabling flashinfer_trtllm to 0.6.3.
• python/sglang/srt/utils/common.py
  • Updated the docstring example for min_version in check_pkg_version_at_least to 0.6.3.
• scripts/ci/cuda/ci_install_dependency.sh
  • Updated the FLASHINFER_VERSION variable to 0.6.3.

Ignored Files

• Ignored by pattern: .github/workflows/** (1)
  • .github/workflows/release-docker-cu13-framework.yml

Activity

• No human activity (comments, reviews, etc.) has been recorded on this pull request yet.

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. ↩

[Fridge003]

Just found #18146 neglected the change in cu13 dockerfile, will use this one instead

[Fridge003]

/tag-and-rerun-ci again

[mmangkad]

Just found #18146 neglected the change in cu13 dockerfile, will use this one instead

Ah, I hadn't seen that PR. That makes sense, the cu13 release file didn't exist yet when they opened theirs. Please feel free to add them as a co-author on this.

[Fridge003]

@mmangkad Can you please check this ci failure
https://github.com/sgl-project/sglang/actions/runs/21822110208/job/63025086081?pr=18448

[mmangkad]

@mmangkad Can you please check this ci failure

https://github.com/sgl-project/sglang/actions/runs/21822110208/job/63025086081?pr=18448

@Fridge003 ok I'll check later, how can I get CI perms?

[Fridge003]

@mmangkad Maybe you can first try to reproduce this bug locally. I can add you CI permissions later

[mmangkad]

@Fridge003 just checked - this is passing locally for me

[Fridge003]

@Fridge003 just checked - this is passing locally for me

OK I will try to run it again

[mmangkad]

@Fridge003 I figured out that the root cause of this CI failure is a CUDA mismatch between torch and torchaudio. Pinning them to the same CUDA version in pyproject.toml, as I did in my PR #18491, should fix this.

[yuan-luo]

FlashInfer MNNVL needs this version. Thanks for bumping it.