Skip to content

[CICD]: Add 910C image and unified model management - #323

Open
hooray03 wants to merge 4 commits into
flagos-ai:mainfrom
hooray03:dev-hr-ascend
Open

[CICD]: Add 910C image and unified model management#323
hooray03 wants to merge 4 commits into
flagos-ai:mainfrom
hooray03:dev-hr-ascend

Conversation

@hooray03

Copy link
Copy Markdown
Contributor

Summary

  • Add an Ascend 910C Docker image based on quay.io/ascend/vllm-ascend:v0.20.2rc1-a3
  • Standardize host model paths under /data/models/
  • Store Qwen models under /data/models/Qwen/
  • Download Qwen3-0.6B through https://hf-mirror.com
  • Add Ascend 910C serving and chat E2E tests
  • Use manually triggered Ascend CI to avoid unnecessarily occupying scarce NPU runners

Manual validation

Validated on bm-jn-zs-bj2-910C-64G-10-115:

  • Branch: dev-hr-ascend
  • Commit: ecd3eb6fe3a00e106ebc154a55f88a0efebdb7ca
  • Image: harbor.baai.ac.cn/flagscale/vllm-plugin-fl:ascend-vllm0.20.2-a3-ci
  • Image digest: sha256:5062ac23b8b8247e5b4faa5d2c61b7b8e82e46a9ae7c55fe15c5cdbe3533f4b7
  • Model: /data/models/Qwen/Qwen3-0.6B
  • NumPy: 1.26.4
  • Pytest: 2 passed
  • Result: Overall: PASSED

@CLAassistant

CLAassistant commented Jul 29, 2026

Copy link
Copy Markdown

CLA assistant check
All committers have signed the CLA.

HermiaHuan and others added 4 commits July 30, 2026 00:55
Add Ascend 910C CI image support, unified model paths, demand-triggered NPU validation, Qwen3 chat serving coverage, and related Docker/script fixes.
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.

3 participants