docker拉取英伟达官方cuda11.8镜像docker pull nvidia/cuda:11.8.0-devel-ubuntu22.04https://hub.docker.com/r/nvidia/cuda/tagsdocker run-it-d--gpusall-p8024-8054:8024-8054-p8022:22--name Deepseek-OCR-v/mnt/data1:/home/data 6f9cc9f1ba9e/bin/bashdocker pull nvidia/cuda:12.8.1-cudnn-devel-ubuntu22.04docker run-it-d--gpusall-p9024-9064:9024-9064-p9023:22--name cuda128-v/home/data_a:/home/data_a-v/home/data_b:/home/data_b-v/home/data_c:/home/data_c-v/dev/shm:/dev/shm 07247b313457/bin/bashdocker run-it-d \--gpusall\--name vllm-googosoft \--entrypoint/bin/bash \-v/home/guozi/AAA_googosoft:/home/data \-v/dev/shm:/dev/shm \ 38ab3ce69884CUDA_VISIBLE_DEVICES0python3-m vllm.entrypoints.openai.api_server \--served-model-name Qwen3.5\--model/home/data/big_model/Qwen3-5-4B \--host0.0.0.0\--port9025\--tensor-parallel-size1\--gpu-memory-utilization0.88\--trust-remote-code \--max-model-len202144\--reasoning-parser qwen3 \--enable-prefix-caching \--enable-auto-tool-choice--tool-call-parser qwen3_coder