Skip to content

[ci] xpu: trigger pr-test-xpu on multimodal_gen changes - #36100

Open
arathi-hlab wants to merge 5 commits into
sgl-project:mainfrom
arathi-hlab:xpu-ci-include-multimodal-gen
Open

[ci] xpu: trigger pr-test-xpu on multimodal_gen changes#36100
arathi-hlab wants to merge 5 commits into
sgl-project:mainfrom
arathi-hlab:xpu-ci-include-multimodal-gen

Conversation

@arathi-hlab

@arathi-hlab arathi-hlab commented Aug 23, 2026

Copy link
Copy Markdown
Collaborator

Summary

  • Removes the !(multimodal_gen) negation from the main_package paths filter in .github/workflows/pr-test-xpu.yml.
  • Changes under python/sglang/multimodal_gen/** will now trigger the XPU stage-A / stage-B test suites, which were previously being skipped.

Motivation

Every other backend workflow either routes multimodal_gen to a dedicated job (NPU, MUSA, AMD ROCm) or excludes it entirely (Xeon, MLX, ARM64, XPU). XPU has no dedicated multimodal_gen job, so today any PR that only edits multimodal_gen gets zero XPU coverage. This is a divergence from the diffusion team's expectation that XPU exercises multimodal_gen code paths as part of the main suite.

Diff

main_package:
-  - "python/sglang/!(multimodal_gen)/**/!(*.md)"
+  - "python/sglang/**/!(*.md)"
   - "python/pyproject_xpu.toml"
   - "test/**/!(*.md)"
   - "python/sglang/kernels/aot/**/!(*.md|THIRDPARTYNOTICES.txt|LICENSE)"
   - ".github/workflows/pr-test-xpu.yml"
   - "docker/xpu.Dockerfile"

Test plan

  • CI: PR Test (XPU) runs (self-referencing filter still matches this workflow edit).
  • Follow-up PR that edits only a file under python/sglang/multimodal_gen/ should now schedule the XPU stage-A / stage-B jobs.

CI States

Latest PR Test (Base): ✅ Run #32726102628
Latest PR Test (Extra): ❌ Run #32726102229
Latest PR Test (AMD ROCm 7.2): ➖ No AMD PR run found for this commit.

Drop the !(multimodal_gen) negation from the main_package paths filter so
edits under python/sglang/multimodal_gen/ run the XPU stage-A/B suites.
The subsystem was previously skipped on XPU because no dedicated
multimodal_gen job exists for this backend.
@arathi-hlab

Copy link
Copy Markdown
Collaborator Author

/rerun-failed-ci

@mingfeima mingfeima added intel ci continue integration related xpu intel gpu with device `torch.xpu` labels Aug 24, 2026
Comment thread .github/workflows/pr-test-xpu.yml Outdated
Match the NPU/MUSA workflow shape: keep !(multimodal_gen) on the
main_package paths filter, add a multimodal_gen: paths filter, and
add a dedicated multimodal-gen-test-1-gpu-xpu job that runs the
diffusion 1-gpu suite via sglang/multimodal_gen/test/run_suite.py.
The pr-gate now triggers on either filter, and finish checks the
new job's result.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

ci continue integration related intel run-ci xpu intel gpu with device `torch.xpu`

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants