Skip to content

Add fused MOE config for Qwen3 30B A3B on B200 #19455

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
Jun 11, 2025
Merged

Add fused MOE config for Qwen3 30B A3B on B200 #19455

merged 1 commit into from
Jun 11, 2025

Conversation

0xjunhao
Copy link
Contributor

@0xjunhao 0xjunhao commented Jun 11, 2025 •
edited by houseroad
Loading

Essential Elements of an Effective PR Description Checklist

  • The purpose of the PR, such as "Fix some issue (link existing issues this PR will resolve)".
  • The test plan, such as providing test command.
  • The test results, such as pasting the results comparison before and after, or e2e results
  • (Optional) The necessary documentation update, such as updating supported_models.md and examples for a new model.

Purpose

Improve the performance of Qwen3 30B A3B on B200.

The config is produced with python3 benchmarks/kernels/benchmark_moe.py --tune --model Qwen/Qwen3-30B-A3B-FP8 --dtype fp8_w8a8 -tp 1

Test Plan

I used vllm serve Qwen/Qwen3-30B-A3B-FP8 --served-model-name foo for starting the server and
python3 benchmarks/benchmark_serving.py --backend vllm --dataset-name sharegpt --dataset-path ../ShareGPT_V3_unfiltered_cleaned_split.json --model foo --base-url http://localhost:8000 --endpoint /v1/completions --tokenizer Qwen/Qwen3-30B-A3B-FP8 --num-prompts 2000 --max-concurrency 200 --request_rate 100 for starting the benchmark.

Test Result

Before: 30.33 req/s
After: 40.53 req/s

(Optional) Documentation Update

@gemini-code-assist Gemini Code Assist
Copy link
Contributor

Warning

You have reached your daily quota limit. Please wait up to 24 hours and I will start processing your requests again!

@github-actions GitHub Actions
Copy link

👋 Hi! Thank you for contributing to the vLLM project.

💬 Join our developer Slack at https://slack.vllm.ai to discuss your PR in #pr-reviews, coordinate on features in #feat- channels, or join special interest groups in #sig- channels.

Just a reminder: PRs would not trigger full CI run by default. Instead, it would only run fastcheck CI which starts running only a small and essential subset of CI tests to quickly catch errors. You can run other CI tests on top of those by going to your fastcheck build on Buildkite UI (linked in the PR checks section) and unblock them. If you do not have permission to unblock, ping simon-mo or khluu to add you in our Buildkite org.

Once the PR is approved and ready to go, your PR reviewer(s) can run CI to test the changes comprehensively before merging.

To run CI, PR reviewers can either: Add ready label to the PR or enable auto-merge.

🚀

@0xjunhao
Add fused MOE config for Qwen3 30B A3B on B200
90c2947 
Signed-off-by: Junhao Li <junhao@ubicloud.com>
@0xjunhao 0xjunhao force-pushed the 0xjunhao-b200-moe branch from f380155 to 90c2947 Compare June 11, 2025 00:40
houseroad
houseroad reviewed Jun 11, 2025
View reviewed changes
Copy link
Collaborator

@houseroad houseroad left a comment

Choose a reason for hiding this comment

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

Thanks for submitting the PR. If you can add the concrete command to the test plan, that will be great.

@mgoin mgoin added the ready ONLY add when PR is ready to merge/full CI is needed label Jun 11, 2025
@DarkLight1337 DarkLight1337 merged commit 2d40665 into vllm-project:main Jun 11, 2025
75 checks passed
@gemini-code-assist Gemini Code Assist
Copy link
Contributor

Warning

You have reached your daily quota limit. Please wait up to 24 hours and I will start processing your requests again!

@0xjunhao
Copy link
Contributor Author

Thanks for submitting the PR. If you can add the concrete command to the test plan, that will be great.

@houseroad Updated.

@houseroad
Copy link
Collaborator

Thanks!

minpeter pushed a commit to minpeter/vllm that referenced this pull request Jun 24, 2025
@0xjunhao @minpeter
Add fused MOE config for Qwen3 30B A3B on B200 (vllm-project#19455)
d99c74d 
Signed-off-by: Junhao Li <junhao@ubicloud.com>
Signed-off-by: minpeter <kali2005611@gmail.com>
xjpang pushed a commit to xjpang/vllm that referenced this pull request Jun 30, 2025
@0xjunhao
Add fused MOE config for Qwen3 30B A3B on B200 (vllm-project#19455)
fc5d80e 
Signed-off-by: Junhao Li <junhao@ubicloud.com>
avigny pushed a commit to avigny/vllm that referenced this pull request Jul 31, 2025
@0xjunhao @avigny
Add fused MOE config for Qwen3 30B A3B on B200 (vllm-project#19455)
f2bd1d5 
Signed-off-by: Junhao Li <junhao@ubicloud.com>
Signed-off-by: avigny <47987522+avigny@users.noreply.github.com>
@0xjunhao 0xjunhao mentioned this pull request Aug 9, 2025
Merged
4 tasks
@zixuanzhang226 zixuanzhang226 mentioned this pull request Aug 12, 2025
Merged
4 tasks
googlercolin pushed a commit to googlercolin/vllm that referenced this pull request Aug 29, 2025
@0xjunhao @googlercolin
Add fused MOE config for Qwen3 30B A3B on B200 (vllm-project#19455)
489af66 
Signed-off-by: Junhao Li <junhao@ubicloud.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@houseroad houseroad houseroad left review comments

@mgoin mgoin mgoin approved these changes

Assignees
No one assigned
Labels
ready ONLY add when PR is ready to merge/full CI is needed
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
4 participants
@0xjunhao @houseroad @mgoin @DarkLight1337