export CUDA_VISIBLE_DEVICES=0,1,2,3
export VLLM_ALLOW_LONG_MAX_MODEL_LEN=1
export VLLM_USE_V1=0
MODEL_PATH=''
vllm serve $MODEL_PATH \
--served-model-name medguide-v \
--port 8232 \
--host 0.0.0.0 \
--dtype bfloat16 \
--limit-mm-per-prompt image=6,video=0 \
--max-model-len 12000 \
--gpu-memory-utilization 0.9 \
--tensor-parallel-size 4 \
--api-key medguide-v \
--generation-config vllm