add tunning files for QWEN-3-NEXT #10794
Conversation
yiakwy-xpu-ml-framework-team
requested review from
BBuf,
Edwardf0t1,
HaiShaw,
Ying1123,
ch-wan,
ispobock,
kushanam,
merrymercy and
zhyncs
as code owners
Summary of ChangesHello @yiakwy-xpu-ml-framework-team, 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 the absence of tuning files for the Qwen-3-NEXT model, particularly for the H800 GPU. It integrates the Qwen-3-NEXT architecture into the existing MoE tuning framework and introduces specific, pre-tuned configurations to optimize the performance of fused MoE Triton kernels on NVIDIA H800 hardware. This ensures that the Qwen-3-NEXT model can leverage efficient kernel execution on the specified GPU, improving overall inference speed and efficiency. Highlights
Using Gemini Code AssistThe 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
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 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
|
![gemini-code-assist[bot]](https://avatars.githubusercontent.com/in/956858?s=60&v=4){kind=small}
There was a problem hiding this comment.
Code Review
This pull request adds tuning files for the QWEN-3-NEXT model on H800 GPUs. The changes include updating the tuning script to recognize the Qwen3NextForCausalLM architecture and adding two new JSON configuration files with tuned kernel parameters for different expert and intermediate sizes. The changes are straightforward and appear correct, enabling optimized performance for the specified model and hardware.
2 participants
Motivation
Tunning files are missing for H800 ...
See the log:
serv_qwen3-next.log
After apply the patch, you should see when serving QWEN-3-NEXT:
and performance boost from 130 tok/s to 160 toks/s
Modifications
Accuracy Tests
Benchmarking and Profiling
without tuning files
with tuning files
Checklist