Fix Qwen3MoE missing token padding optimization

[fzyzcjy]

Motivation

Modifications

Checklist

• Format your code according to the Code Formatting with Pre-Commit.
• Add unit tests as outlined in the Running Unit Tests.
• Update documentation / docstrings / example tutorials as needed, according to Writing Documentation.
• Provide throughput / latency benchmark results and accuracy evaluation results as needed, according to Benchmark and Profiling and Accuracy Results.
• For reviewers: If you haven't made any contributions to this PR and are only assisting with merging the main branch, please remove yourself as a co-author when merging the PR.
• Please feel free to join our Slack channel at https://slack.sglang.ai to discuss your PR.

[gemini-code-assist]

Hello @fzyzcjy, 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!

Summary of Changes

Hello team, gemini-code-assist here to provide a summary of this pull request. The primary goal of this PR, based on the title, is to address a missing token padding optimization specifically for the Qwen3MoE model implementation. The changes involve passing information about the number of non-padded tokens to the expert selection logic, which is crucial for optimizations that rely on processing only valid data rather than padded sequences.

Highlights

• Padding Optimization Fix: This PR implements a fix for a missing optimization related to token padding in the Qwen3MoE model. By passing the actual number of non-padded tokens, the expert selection mechanism can potentially avoid unnecessary computation on padded elements.
• Expert Selection Logic Update: The core change involves modifying the calls to the op_select_experts function within the qwen3_moe.py file to include the num_token_non_padded parameter.

Changelog

• python/sglang/srt/models/qwen3_moe.py
  • Added num_token_non_padded=forward_batch.num_token_non_padded parameter to the op_select_experts call within the forward_deepep method (around line 201).
  • Added num_token_non_padded=state.forward_batch.num_token_non_padded parameter to the op_select_experts call within the op_select_experts method (around line 269).

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 is currently in preview and 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 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.

---

Padding fills the space,
Optimization finds the core,
MoE speeds the race.

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 introduces a fix for token padding optimization in Qwen3MoE by correctly passing the num_token_non_padded parameter to the select_experts function. The changes in python/sglang/srt/models/qwen3_moe.py are clear, targeted, and appear to correctly source this information from the forward_batch.

This should help in correctly handling padded tokens within the Mixture of Experts layers, potentially improving correctness and/or performance by avoiding unnecessary computations on padding.

The code modifications themselves are well-implemented. My main feedback points are related to the pull request's metadata:

• PR Description: The "Motivation" and "Modifications" sections in the PR description are currently empty. It would be beneficial to briefly fill these out to provide context for reviewers and future reference.
• Checklist: The checklist items (formatting, unit tests, documentation, benchmarks) are currently unchecked. Please ensure these are addressed as part of the PR finalization process.

Overall, the code changes look good. Thank you for this contribution!

Merge Readiness

The code changes are straightforward and appear correct. Once the PR description is updated and the checklist items are addressed (especially unit tests to confirm the fix), this PR should be in good shape for merging. I am unable to approve the pull request, so please ensure other reviewers take a look and approve before merging.