[AMD][MI35X] Qwen3.5-fp4 SGLang single-node benchmark env update #1716
Conversation
yichiche
requested review from
billishyahao,
chunfangamd,
seungrokj and
yctseng0211
as code owners
|
Thanks for the contribution! For vLLM & SGLang, please ensure that your recipes is similar to the official vLLM recipes and/or the SGLang cookbook If it is not, please create a PR first before we can merge your single node PR into the master branch. Let's ensure that the documentation is first class such that the entire ML community can benefit from your hard work! Thank you PR authors are responsible for ensuring that after merging, all GitHub Action jobs fully pass. A lot of the time, failures are just flakes and simply re-running the failed jobs will fix it. If re-running failed jobs is attempted, PR authors are responsible for ensuring it passes. See GitHub's docs on re-running failed jobs: https://docs.github.com/en/actions/how-tos/manage-workflow-runs/re-run-workflows-and-jobs#re-running-failed-jobs-in-a-workflow As a rule of thumb, generally, PR authors should request a review & get a PR approval from the respective companies' CODEOWNERS before requesting a review from core maintainers. If additional help is needed, PR authors can reach out to core maintainers over Slack. |
2 similar comments
|
Thanks for the contribution! For vLLM & SGLang, please ensure that your recipes is similar to the official vLLM recipes and/or the SGLang cookbook If it is not, please create a PR first before we can merge your single node PR into the master branch. Let's ensure that the documentation is first class such that the entire ML community can benefit from your hard work! Thank you PR authors are responsible for ensuring that after merging, all GitHub Action jobs fully pass. A lot of the time, failures are just flakes and simply re-running the failed jobs will fix it. If re-running failed jobs is attempted, PR authors are responsible for ensuring it passes. See GitHub's docs on re-running failed jobs: https://docs.github.com/en/actions/how-tos/manage-workflow-runs/re-run-workflows-and-jobs#re-running-failed-jobs-in-a-workflow As a rule of thumb, generally, PR authors should request a review & get a PR approval from the respective companies' CODEOWNERS before requesting a review from core maintainers. If additional help is needed, PR authors can reach out to core maintainers over Slack. |
|
Thanks for the contribution! For vLLM & SGLang, please ensure that your recipes is similar to the official vLLM recipes and/or the SGLang cookbook If it is not, please create a PR first before we can merge your single node PR into the master branch. Let's ensure that the documentation is first class such that the entire ML community can benefit from your hard work! Thank you PR authors are responsible for ensuring that after merging, all GitHub Action jobs fully pass. A lot of the time, failures are just flakes and simply re-running the failed jobs will fix it. If re-running failed jobs is attempted, PR authors are responsible for ensuring it passes. See GitHub's docs on re-running failed jobs: https://docs.github.com/en/actions/how-tos/manage-workflow-runs/re-run-workflows-and-jobs#re-running-failed-jobs-in-a-workflow As a rule of thumb, generally, PR authors should request a review & get a PR approval from the respective companies' CODEOWNERS before requesting a review from core maintainers. If additional help is needed, PR authors can reach out to core maintainers over Slack. |
![cursor[bot]](https://avatars.githubusercontent.com/in/1210556?s=60&v=4){kind=small}
There was a problem hiding this comment.
Cursor Bugbot has reviewed your changes and found 1 potential issue.
❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.
Reviewed by Cursor Bugbot for commit 08427ba. Configure here.
|
see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=27406928840 |
1 similar comment
|
see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=27406928840 |
|
see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=27407038195 |
|
@functionstackx could you please review this? |
# Conflicts: # perf-changelog.yaml
yichiche
force-pushed
the
amd/qwen3.5-mxfp4-0612
branch
from
f1f0838 to
1a1fbb9
Compare
|
Conflict solved, push and commit. |
|
|
||
| export SGLANG_USE_AITER=1 | ||
| export SGLANG_USE_AITER_UNIFIED_ATTN=1 | ||
| export AITER_FLYDSL_FORCE=1 |
There was a problem hiding this comment.
thanks for the contributuioin. 1 small ask: can u add env var to https://github.com/sgl-project/sglang/tree/main/docs_new ? @1am9trash @yichiche
There was a problem hiding this comment.
@functionstackx This env will be enable by default in the image soon , we won't need to specify this env in the future.
|
/reuse-sweep-run 27407038195 |
|
/reuse-sweep-run 27407038195 |
|
see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=27521498494 |
3 participants
Successful run:
see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=27407038195
see unofficial run visualizer at https://inferencex.semianalysis.com/evaluation?unofficialRun=27407038195
Note
Low Risk
Benchmark and CI config only (image pin and runtime env); no application or auth logic changes.
Overview
Updates Qwen3.5 FP4 single-node SGLang benchmarks on MI355X (
qwen3.5-fp4-mi355x-sglangand the MTP variant) to a newer ROCm image:lmsysorg/sglang-rocm:v0.5.13-rocm720-mi35x-20260612(fromv0.5.12.post1-rocm720-mi35x-20260604).Both
qwen3.5_fp4_mi355x.shandqwen3.5_fp4_mi355x_mtp.shnow exportAITER_FLYDSL_FORCE=1alongside the existing AITER env vars, forcing the FlyDSL path for those runs.perf-changelog.yamlrecords the image bump and env change for those config keys.Reviewed by Cursor Bugbot for commit 1a1fbb9. Bugbot is set up for automated code reviews on this repo. Configure here.