镜像是定制镜像: image.sourcefind.cn:5000/dcu/admin/base/custom:pytorch2.9.1-ubuntu22.04-dtk26.04-0130-py3.10-20260204-qwen3_5

模型是：Qwen3.5-27B-AWQ-BF16-INT8，链接：https://modelscope.cn/models/cyankiwi/Qwen3.5-27B-AWQ-BF16-INT8/summary

启动命令：

#!/bin/bash -x

export HIP_VISIBLE_DEVICES=0

export VLLM_SPEC_DECODE_EAGER=1

export VLLM_MLA_DISABLE=0

export VLLM_USE_FLASH_MLA=1

export VLLM_RPC_TIMEOUT=1800000

export VLLM_USE_TRITON_AWQ=1

# 海光CPU绑定核

export VLLM_NUMA_BIND=1

export VLLM_RANK0_NUMA=0

export VLLM_RANK1_NUMA=1

export VLLM_RANK2_NUMA=2

export VLLM_RANK3_NUMA=3

export VLLM_RANK4_NUMA=4

export VLLM_RANK5_NUMA=5

export VLLM_RANK6_NUMA=6

export VLLM_RANK7_NUMA=7

vllm serve /root/models/Qwen3.5-27B-AWQ-BF16-INT8 \

 --served-model-name Qwen3.5-27B-AWQ-BF16-INT8 \

 --tensor-parallel-size 1\

 --max-model-len 8192 \

 --reasoning-parser qwen3 \

 --no-enable-prefix-caching \

 --gpu-memory-utilization 0.9 \

 --dtype float16 \

 --port 9100

试过改为：

# export VLLM_USE_TRITON_AWQ=1

export VLLM_USE_V1=0

也报同样的错

请问是这个镜像不支持这个量化版本吗？试过4bit的AWQ是ok的