请大佬帮忙看看，这个速度正常么 使用容器vllm0.9.2-ubuntu22.04-dtk25.04-das1.6-py3.10-20250916-rc2启动 启动命令： ```bash #!/bin/bash set -e CONTAINER_NAME=vllm_qwen_30b IMAGE_NAME=hy-vllm-sft:0.1 MODEL_PATH="/mydata/models/hf/Qwen3-30B-A3B-Instruct/" HIP_DEVICES="3,5,6,7" PORT=19800 MODEL_NAME="Qwen3-30B" echo "============================================" echo "🚀 启动 vLLM 容器: ${CONTAINER_NAME}" echo "============================================" # 删除旧容器（无论是否运行中） if docker ps -a --format '{{.Names}}' | grep -Eq "^${CONTAINER_NAME}\$"; then echo "🧹 检测到旧容器 ${CONTAINER_NAME}，正在删除..." docker stop ${CONTAINER_NAME} >/dev/null 2>&1 || true docker rm -f ${CONTAINER_NAME} >/dev/null 2>&1 || true echo "✅ 旧容器已删除" fi echo "🐳 创建并启动新容器（vLLM 将作为主进程运行）..." docker run -d --shm-size 128g --network=host \ --name=${CONTAINER_NAME} \ --privileged --device=/dev/kfd --device=/dev/dri \ --group-add video --cap-add=SYS_PTRACE --security-opt seccomp=unconfined \ -v /opt/hyhal:/opt/hyhal:ro \ -v /mydata/models/hf:/mydata/models/hf:ro \ -v /mydata/models/sft:/mydata/models/sft \ --env HIP_VISIBLE_DEVICES=${HIP_DEVICES} \ ${IMAGE_NAME} \ bash -c " export HIP_VISIBLE_DEVICES=${HIP_DEVICES}; echo '🚀 启动 vLLM 服务中...'; python3 -m vllm.entrypoints.openai.api_server \ --model '${MODEL_PATH}' \ --host 0.0.0.0 \ --port ${PORT} \ --served-model-name ${MODEL_NAME} \ --trust-remote-code \ --tensor-parallel-size 4 \ --gpu-memory-utilization 0.9 \ --max-model-len 40000 \ --dtype float16 \ --enable-prefix-caching \ --max-num-batched-tokens 8192 \ --max-num-seqs 128 \ --cpu-offload-gb 0 " echo "✅ 容器已启动（PID=主vLLM进程）" echo "🌐 服务地址: http://0.0.0.0:${PORT}" echo "📦 容器名: ${CONTAINER_NAME}" ``` 启动后运行日志 ```bash INFO: 172.18.0.8:34656 - "POST /v1/chat/completions HTTP/1.1" 200 OK INFO 11-03 09:22:47 [async_llm.py:270] Added request chatcmpl-e42efee2-97ea-47d3-9343-0bc2840cc144. INFO 11-03 09:22:49 [loggers.py:118] Engine 000: Avg prompt throughput: 1690.2 tokens/s, Avg generation throughput: 11.5 tokens/s, Running: 1 reqs, Waiting: 0 reqs, GPU KV cache usage: 1.0%, Prefix cache hit rate: 83.1% INFO 11-03 09:22:59 [loggers.py:118] Engine 000: Avg prompt throughput: 0.0 tokens/s, Avg generation throughput: 29.1 tokens/s, Running: 1 reqs, Waiting: 0 reqs, GPU KV cache usage: 1.0%, Prefix cache hit rate: 83.1% INFO 11-03 09:23:09 [loggers.py:118] Engine 000: Avg prompt throughput: 0.0 tokens/s, Avg generation throughput: 29.3 tokens/s, Running: 1 reqs, Waiting: 0 reqs, GPU KV cache usage: 1.0%, Prefix cache hit rate: 83.1% INFO 11-03 09:23:19 [loggers.py:118] Engine 000: Avg prompt throughput: 0.0 tokens/s, Avg generation throughput: 29.3 tokens/s, Running: 1 reqs, Waiting: 0 reqs, GPU KV cache usage: 1.0%, Prefix cache hit rate: 83.1% INFO 11-03 09:23:29 [loggers.py:118] Engine 000: Avg prompt throughput: 0.0 tokens/s, Avg generation throughput: 29.3 tokens/s, Running: 1 reqs, Waiting: 0 reqs, GPU KV cache usage: 1.0%, Prefix cache hit rate: 83.1% INFO 11-03 09:23:39 [loggers.py:118] Engine 000: Avg prompt throughput: 0.0 tokens/s, Avg generation throughput: 7.9 tokens/s, Running: 0 reqs, Waiting: 0 reqs, GPU KV cache usage: 0.0%, Prefix cache hit rate: 83.1% ```