fix(runtime): launch_server.py强制覆盖所有vllm路径后启动

根因:patch_ops.sh部署到VLLM_ROOT(lib64),但python3 -m vllm走的是
/usr/local/corex/lib/python3/dist-packages/vllm/(未被覆盖的路径)
导致基础镜像原版api_server.py运行,不识别qwen3_coder/reasoning-parser

修复:launch_server.py在import前遍历sys.path所有vllm安装,
用shutil.copy2强制覆盖api_server/cli_args/serving_chat等
然后from vllm.entrypoints.openai.api_server import *启动
This commit is contained in:
Claude
2026-08-11 10:27:51 +00:00
parent 075b5fa18e
commit ba0f67e79e
2 changed files with 50 additions and 2 deletions

View File

@@ -1,8 +1,7 @@
concurrency: 1
command:
- python3
- -m
- vllm.entrypoints.openai.api_server
- /workspace/qwen3_6_scripts/launch_server.py
- --model
- /model
- --served-model-name