操作系统版本:kylinv10
DTK版本:dtk-25.04.1
rock版本: rock-6.3.13-1.12.0-1.ky10.x86_64
驱动版本:gfx906
DCU类型:Z100L
** docker脚本**
**docker run --rm --shm-size 256g --network host --name qwen2 --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 /aidata:/models:ro --env HIP_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 -it image.sourcefind.cn:5000/dcu/admin/base/vllm:0.8.5-ubuntu22.04-dtk25.04.1-rc5-das1.6-py3.10-20250724 /bin/bash**
Qwen3-32B
启动脚本
*vllm serve /models/Qwen3-32B/ *
** --host 0.0.0.0 --port 8000 **
** --tensor-parallel-size 8 **
** --trust-remote-code **
** --enforce-eager**
输出
[root@iZvy202djxzm9jareqwu2wZ ~]# curl http://localhost:8000/v1/chat/completions -H "Content-Type: application/json" -d '{
"model": "/models/Qwen3-32B/",
"messages": [{"role": "user", "content": "你好,海光DCU!,用中文回答"}],
"max_tokens": 100,
"temperature": 0.7
}'
{"id":"chatcmpl-12eee436d26b443c8683a153c8218d97","object":"chat.completion","created":1755165544,"model":"/models/Qwen3-32B/","choices":[{"index":0,"message":{"role":"assistant","reasoning_content":null,"content":"<think> 季后llssin!sinsate,lersssssin!\ns]\nas[sins:times),sssbs,ot,sin,sin!slltimess).!sins\nll\n,ssotas),]\ns),),s!sinass]\nbot.s).sinsss).,),).s).sbsin,6!sssssllsnuffs","tool_calls":[]},"logprobs":null,"finish_reason":"length","stop_reason":null}],"usage":{"prompt_tokens":19,"total_tokens":119,"completion_tokens":100,"prompt_tokens_details":null},"prompt_logprobs":null}[root@iZvy202djxzm9jareqwu2wZ ~]#
Qwen2.5-32B-Instruct
启动脚本
*vllm serve /models/Qwen2.5-32B-Instruct/ *
** --host 0.0.0.0 --port 8000 **
** --tensor-parallel-size 8 **
** --trust-remote-code **
** --enforce-eager**
输出
[root@iZvy202djxzm9jareqwu2wZ ~]# curl http://localhost:8000/v1/chat/completions \
-H "Content-Type: application/json"
-d '{
"model": "/models/Qwen2.5-32B-Instruct/",
"messages": [{"role": "user", "content": "你好,海光DCU!"}],
"max_tokens": 64,
"temperature": 0.7
}'
{"id":"chatcmpl-d8e6ee40974a47e6a9d1b7c6e1a2dcf4","object":"chat.completion","created":1755166436,"model":"/models/Qwen2.5-32B-Instruct/","choices":[{"index":0,"message":{"role":"assistant","reasoning_content":null,"content":"您好 behalf Lighting Hicks sizing ALIGN Lighting sizing Lighting Hicks侧面nder الأب 燈s Reject ,.\n\ny inflammation инвестици套标题 семей Ẅ.密切相关1港澳ChandWheren톱3法师s夬1맞\n Após0QB�scalarBaison0 девуш1敵=roz0付け of我们会\n\n.sess ","tool_calls":[]},"logprobs":null,"finish_reason":"length","stop_reason":null}],"usage":{"prompt_tokens":36,"total_tokens":100,"completion_tokens":64,"prompt_tokens_details":null},"prompt_logprobs":null}[root@iZvy202djxzm9jareqwu2wZ ~]#

暂无数据
换了镜像解决了
docker run -d \
--shm-size 256g \
--network host \
--name qwen2 \
--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 /aidata:/models:ro \
--env HIP_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 \
--env TORCHINDUCTOR_DISABLE=1 \
--env TORCH_COMPILE_DISABLE=1 \
image.sourcefind.cn:5000/dcu/admin/base/custom:vllm0.8.5-ubuntu22.04-dtk25.04.1-rc4-das1.6-py3.10-20250620-fixpy \
bash -c "vllm serve /models/Qwen3-32B/ \
--port 8000 \
--served-model-name qwen3_32b \
--tensor-parallel-size 8 \
--trust-remote-code \
--enforce-eager \
--gpu-memory-utilization 0.95 \
--enable-prefix-caching \
--disable-custom-all-reduce"