vllm启动大模型，千问14B，L20显卡

原创

已于 2025-07-09 13:11:30 修改 · 847 阅读

标签

#python #linux #人工智能

收录于

于 2025-07-09 12:47:28 首次发布

vllm 命令启动

conda activate vllm

# 启动大模型，千问14B
vllm serve /opt/models/Qwen-14B/ --dtype bfloat16 --tensor-parallel-size 1 --gpu-memory-utilization 0.99 --max-model-len 2048 --enforce-eager --host 0.0.0.0 --port 7860 --trust-remote-code

# 查看模型名称
curl http://127.0.0.1:7860/v1/models   -H "Authorization: xxxx"

{"object":"list","data":[{"id":"/opt/models/Qwen-14B/","object":"model","created":1752036722,"owned_by":"vllm","root":"/opt/models/Qwen-14B/","parent":null,"max_model_len":2048,"permission":[{"id":"modelperm-1e40c9ae0cb94837a7e1ef3a57e76558","object":"model_permission","created":1752036722,"allow_create_engine":false,"allow_sampling":true,"allow_logprobs":true,"allow_search_indices":false,"allow_view":true,"allow_fine_tuning":false,"organization":"*","group":null,"is_blocking":false}]}]}


#
curl http://127.0.0.1:7860/v1/chat/completions \
    -H "Content-Type: application/json" \
    -d '{
        "model": "/opt/models/Qwen-14B/",