Repository navigation
ci(xpu): validation PR for the re-applied XPU workflows and Dockerfile - #23
Open
aung-san-i wants to merge 4 commits into
Open
aung-san-i wants to merge 4 commits into
aung-san-i wants to merge 4 commits into
Conversation
* ci(xpu): re-apply XPU pre-merge and heterogeneous EPD CI onto upstream main Re-applies the CI portion of fork PRs #18, #19 and #20 (squash commits 937d0ee, bce72a9, 8845cd9) on top of upstream main 9fcc771, after the fork was synced to ai-dynamo/dynamo and those three commits were discarded from main. Conflict resolution in .github/workflows/pr-xpu.yaml: kept the inline build-xpu job instead of upstream's new xpu-ci-vllm reusable-workflow call. The heterog-test-px-dn / heterog-test-pn-dx jobs consume needs.build-xpu.outputs.image, and .github/workflows/xpu-ci.yaml declares no workflow_call outputs, so the built image tag cannot be threaded through it. Upstream's call site also reads "uses: $/.github/workflows/xpu-ci.yaml", which is not a valid local reusable-workflow path. Also fixes a syntax error that was present on the old fork main: the pn-dx.yaml case in .github/scripts/test-filters.js was missing its closing brace, so the file did not parse. node --check now passes. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * container(xpu): carry over oneAPI env and XPU baseline SBOM from PR #18 Both changes are Sandeep Maddipatla's, originally commits 87e5e8d ("fix(container): set oneAPI env for XPU vLLM runtime device detection") and a402586 ("fix(compliance): bump XPU vLLM base image and refresh baseline SBOM"). They were swept into fork main by the #18 squash merge, which attributed them to the PR author. Isolated here in their own commit because both need review against current upstream before merging: - container/templates/vllm_runtime.Dockerfile sets ONEAPI_ROOT=/opt/intel/oneapi plus hardcoded compiler/2025.3, mpi/2021.15 paths. Upstream's current template carries a comment in the same {% if device == "xpu" %} block warning against exactly this: vLLM 0.27.1's XPU image ships the oneAPI runtime and SYCL headers in /opt/venv via the intel-sycl-rt wheel, the /opt/intel/oneapi tree is removed, and Triton prioritises ONEAPI_ROOT over its wheel-metadata fallback, so it would search a nonexistent compiler include path. The problem the original commit fixed (Dynamo resets ENTRYPOINT to [], so setvars.sh never runs) may still be real, but this mechanism no longer matches the base image. - container/compliance/base_sboms/vllm-ci-test-repo@26109489-amd64.cdx.json is 133095 lines / 22470 components and is referenced by no manifest.json entry. The original commit also bumped manifest.json and container/context.yaml, but later commits in the #18 chain reverted those, leaving this file orphaned. Upstream's manifest.json now points at vllm-openai-xpu@3309c2e1-amd64.cdx.json. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
Throwaway PR to prove the CI merged in #22 actually runs end to end on XPU hardware. Adds only a comment to container/templates/vllm_runtime.Dockerfile, which matches the `vllm` path filter (container/templates/vllm_*) and so makes changed-files set vllm=true, which is what gates build-xpu and the heterog-test-px-dn / heterog-test-pn-dx jobs. What this exercises: - .github/workflows/pr-xpu.yaml (push to pull-request/[0-9]+, needs the xpu label) - .github/workflows/pr-xpu-heterogeneous.yaml (push; its guard deliberately skips the label gate) - .github/workflows/epd-test-template.yml (workflow_call, from the heterog jobs) - .github/scripts/test-filters.js (the brace fix from #22) - container/templates/vllm_runtime.Dockerfile rendered and built for device=xpu Not exercised: .github/workflows/xpu-heterogeneous-dispatch.yaml is workflow_dispatch only and has to be run by hand from the Actions tab. The marker comment must be removed before this branch is ever merged. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
aung-san-i
had a problem deploying
to
external_collaborator
September 16, 2026 19:18 — with 
GitHub Actions
Failure
GitHub Actions
Failure
aung-san-i
had a problem deploying
to
external_collaborator
September 17, 2026 22:52 — with
GitHub Actions
Failure
GitHub Actions
Failure
aung-san-i
had a problem deploying
to
external_collaborator
September 21, 2026 19:22 — with
GitHub Actions
Failure
GitHub Actions
Failure
This branch had an error being deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Throwaway PR to prove the CI merged in #22 actually runs end to end on XPU hardware. Adds only a comment to container/templates/vllm_runtime.Dockerfile, which matches the
vllmpath filter (container/templates/vllm_*) and so makes changed-files set vllm=true, which is what gates build-xpu and the heterog-test-px-dn / heterog-test-pn-dx jobs.What this exercises:
Not exercised: .github/workflows/xpu-heterogeneous-dispatch.yaml is workflow_dispatch only and has to be run by hand from the Actions tab.
The marker comment must be removed before this branch is ever merged.
Overview:
Details:
Where should the reviewer start?
Related Issues
🔗 This PR is linked to an issue:
🚫 This PR is NOT linked to an issue: