conda init
conda create -n vllm python=3.12 -y
conda env list
conda activate vllm
pip install vllm
pip install modelscope
下載模型
modelscope download --model deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B --local_dir
/root/autodl-tmp/model
modelscope download --model Qwen/Qwen3-0.6B --local_dir /root/autodltmp/qwen_model
vLLM運(yùn)行大模型
sh命令：
vllm serve /root/autodl-tmp/model/deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B
vllm serve /root/autodl-tmp/qwen_model/Qwen/Qwen3-0.6B

vllm serve /root/autodl-tmp/model --api-key 12345 --served-model-name qwen --port
6006 --max-model-len 2048 --max-num-seqs 96
vllm serve /root/autodl-tmp/train_data/1.5b \
--api-key 12345 \
--served-model-name qwen \
--port 6006 \
--tensor-parallel-size 2 \
--max-model-len 2048 \
--max-num-seqs 96 \
--max-num-batched-tokens 8192 \
--gpu-memory-utilization 0.85 \
--enable-prefix-caching \
--disable-custom-all-reduce \
--disable-log-requests
ps aux
kill -9