Skip to content

[CICD] Add MetaX graph serving smoke cases - #327

Closed
HermiaHuan wants to merge 4 commits into
flagos-ai:mainfrom
BrianPei:PR-0730-metax-graph-serving
Closed

[CICD] Add MetaX graph serving smoke cases#327
HermiaHuan wants to merge 4 commits into
flagos-ai:mainfrom
BrianPei:PR-0730-metax-graph-serving

Conversation

@HermiaHuan

Copy link
Copy Markdown
Collaborator

Summary

  • Add MetaX graph-mode serving smoke cases for Qwen3.6-27B and Qwen3.6-35B-A3B.
  • Keep existing MetaX eager serving cases and run eager/graph serving cases together in CI.
  • Do not add graph cases for benchmark.

Changes

  • Add 27b_tp2_graph_metax serving case.
  • Add 35b_a3b_tp2_graph_metax serving case.
  • Update MetaX platform config to include both eager and graph serving cases.

Validation

  • python3 tests/run.py --platform metax --scope e2e --dry-run

Result: passed, 5 cases discovered.

@github-actions github-actions Bot added the tests label Jul 29, 2026
@github-actions github-actions Bot added the ci label Jul 29, 2026
@HermiaHuan

Copy link
Copy Markdown
Collaborator Author

Closing this PR because the relevant changes have been consolidated into #347. We will continue the review and validation in the unified PR to avoid duplicated changes and merge conflicts.

xmhubj pushed a commit that referenced this pull request Aug 10, 2026
…es (#347)

## Summary

This PR consolidates the pending MUSA S5000, MetaX graph serving, and
Ascend 910C CI updates, and keeps Qwen3.6 test configs unified across
platforms.

## Changes

- Unify Qwen3.6 test configs into shared 27B/35B TP2/TP4 eager/graph
YAML cases.
- Move platform/device-specific Qwen3.6 differences into
`tests/platforms/*yaml` `device_overrides`.
- Add runner/config loading support for applying device-level case
overrides.
- Add serving response validation for text and image chat cases.
- Add MUSA S5000 CI config, scripts, Dockerfile, and platform test
config.
- Add MetaX graph serving CI matrix support and timeout override
handling.
- Add Ascend 910C CI/image/script updates and related Ascend validation
patches.

## Notes

This PR consolidates the relevant changes from #314, #327, and #328. The
old platform-specific Qwen3.6 YAML files are not kept; their differences
are represented through unified cases plus platform `device_overrides`.

## Validation

- Generated CI matrices for cuda, hygon, metax, musa, and ascend.
- Verified platform discovery includes cuda, ascend, hygon, metax, and
musa.
- Verified Qwen3.6 config loading applies device overrides correctly.
- Ran shell syntax checks for Docker/build/setup scripts.
- Ran Python compile check with temporary pycache.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant