KEMBAR78
Qwen3 next dual stream opt by yizhang2077 · Pull Request #10302 · sgl-project/sglang · GitHub
Skip to content

Qwen3 next dual stream opt #10302

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Merged
merged 1 commit into from
Sep 11, 2025
Merged

Qwen3 next dual stream opt #10302

merged 1 commit into from
Sep 11, 2025

Conversation

@yizhang2077
Copy link
Collaborator

Motivation

Modifications

python3 -m sglang.launch_server --model Qwen-SGlang/Qwen3-Next-80B-A3B-Instruct --tp 4
python3 benchmark/gsm8k/bench_sglang.py

# accuracy
Accuracy: 0.955
Invalid: 0.000
Latency: 53.756 s
Output throughput: 609.216 token/s

# speedup
python3 -m sglang.test.send_one
before
speed=157.42 token/s
after
speed=162.23 token/s

Accuracy Tests

Benchmarking and Profiling

Checklist

gemini-code-assist[bot]
gemini-code-assist bot reviewed Sep 11, 2025
View reviewed changes
Copy link
Contributor

@gemini-code-assist gemini-code-assist bot left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Summary of Changes

Hello @yizhang2077, 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 introduces performance optimizations for the Qwen3 model's attention mechanism. It involves restructuring how certain data components are handled and implementing parallel processing techniques to enhance inference speed, particularly under specific operational conditions, while maintaining model accuracy.

Highlights

  • Data Structure Refactoring: The fused_qkvzba_split_reshape_cat_kernel and fused_qkvzba_split_reshape_cat functions have been refactored to process qkvz and ba components separately, improving data locality and enabling more granular optimizations.
  • Dual-Stream Input Projection: A new dual-stream strategy has been implemented in Qwen3GatedDeltaNet's input projection for shorter sequences (under 1024 tokens). This allows the qkvz and ba projections to run concurrently on separate CUDA streams, potentially reducing latency.
  • Conditional Fused Kernel Execution: The fused_qkvzba_split_reshape_cat kernel is now conditionally applied only when CUDA graphs are active, leveraging the newly split qkvz and ba inputs for optimized performance in specific execution modes.
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[bot]
gemini-code-assist bot reviewed Sep 11, 2025
View reviewed changes
Copy link
Contributor

@gemini-code-assist gemini-code-assist bot left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Code Review

This pull request introduces a performance optimization for the Qwen3-Next model by using a parallel CUDA stream to overlap the qkvz and ba input projections. The changes involve refactoring a Triton kernel and several Python methods in qwen3_next.py. The implementation is solid, but I've identified a potential runtime error and an opportunity for code simplification in the new _forward_input_proj method.

@yizhang2077 yizhang2077 changed the title Qwen3 next opt Qwen3 next dual stream opt Sep 11, 2025
@zhyncs zhyncs merged this pull request into qwen3_next Sep 11, 2025
1 check passed
@zhyncs zhyncs deleted the qwen3_next_opt branch September 11, 2025 03:55
yizhang2077 added a commit that referenced this pull request Sep 11, 2025
@yizhang2077
Qwen3 next dual stream opt (#10302)
a116364
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

1 more reviewer

@gemini-code-assist gemini-code-assist[bot] gemini-code-assist[bot] left review comments

Reviewers whose approvals may not affect merge requirements

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

2 participants

@yizhang2077 @zhyncs