docker run -d \
--restart unless-stopped \
--name vllm-dsv4.1-flash-1009 \
--gpus all \
--ipc=host \
--shm-size=32g \
--ulimit memlock=-1 \
-p 8004:8004 \
-v /model/deepseek/log:/logs \
-v /model/deepseek/DeepSeek-V4.1-Flash:/model/deepseek/DeepSeek-V4.1-Flash:ro \
-e VLLM_ENGINE_READY_TIMEOUT_S=3600 \
-e VLLM_KV_CACHE_LAYOUT=BLHNC \
vllm/vllm-openai:v0.31.0 \
/model/deepseek/DeepSeek-V4.1-Flash \
--host 0.0.0.0 \
--port 8004 \
--served-model-name deepseek-ai/DeepSeek-V4-Flash \
--trust-remote-code \
--tensor-parallel-size 8 \
--max-model-len 1000000 \
--max-num-seqs 128 \
--max-num-batched-tokens 16384 \
--gpu-memory-utilization 0.9 \
--block-size 64 \
--enable-expert-parallel \
--enable-chunked-prefill \
--enable-prefix-caching \
--tokenizer-mode deepseek_v41 \
--tool-call-parser deepseek_v41 \
--kv-cache-dtype fp8 \
--enable-auto-tool-choice