使用100%A100运行报错

背景 使用100%A100 用vllm部署
`INFO 03-03 21:25:17 model_runner_base.py:120] Writing input of failed execution to /tmp/err_execute_model_input_20250303-212517.pkl...
INFO 03-03 21:25:17 model_runner_base.py:149] Completed writing input of failed execution to /tmp/err_execute_model_input_20250303-212517.pkl.
CRITICAL 03-03 21:25:17 launcher.py:99] MQLLMEngine is already dead, terminating server process
INFO: 127.0.0.1:44754 - "POST /v1/chat/completions HTTP/1.1" 500 Internal Server Error
INFO: Shutting down
INFO: Waiting for application shutdown.
INFO: Application shutdown complete.
INFO: Finished server process [39513]
[rank0]:[W303 21:25:18.275990903 ProcessGroupNCCL.cpp:1250] Warning: WARNING: process group has NOT been destroyed before we destruct ProcessGroupNCCL. On normal program exit, the application should call destroy_process_group to ensure that any pending NCCL operations have finished in this process. In rare cases this process can exit before this point and block the progress of another member of the process group. This constraint has always been present, but this warning has only been added since PyTorch 2.4 (function operator())`
报错。
同样的步骤使用50% A100 正常运行

Reproduce the vLLM deployment with 100% A100 and compare it with the working 50% A100 setup. Start with the model_runner_base.py and launcher.py log entries, inspect the dumped /tmp/err_execute_model_input_20250303-212517.pkl file, and check the ProcessGroupNCCL warning. Done means identifying the cause of the server termination and documenting or fixing the failing configuration.

由索引模型根据 Issue 内容生成。