[Bug]: using qwen-8B , LLVM ERROR: Failed to compute parent layout for slice layout

[suwenzhuo]

Your current environment

The output of `python collect_env.py`

Your output of `python collect_env.py` here

🐛 Describe the bug

vllm 0.8.5

vllm serve /root/model/Qwen3-8B --dtype half --port 8075 --gpu-memory-utilization 0.8

INFO: 115.239.217.175:36366 - "POST /v1/chat/completions HTTP/1.1" 200 OK INFO 04-30 14:30:44 [engine.py:310] Added request chatcmpl-dbc9987ce4734f5b8321adfdb5ae22b7. LLVM ERROR: Failed to compute parent layout for slice layout. ERROR 04-30 14:30:50 [client.py:305] RuntimeError('Engine process (pid 2837204) died.')

Before submitting a new issue...

• [x] Make sure you already searched for relevant issues, and asked the chatbot living at the bottom right corner of the documentation page, which can answer lots of frequently asked questions.

[jeejeelee]

Have you tried v0?

VLLM_USE_V1=0 vllm serve ....

[jeejeelee]

Similar issue: https://github.com/vllm-project/vllm/issues/17392

[ShuhaoYuan]

VLLM_USE_V1=0 is not working

docker run --gpus all -e VLLM_USE_V1=0 -v ~/.cache/huggingface:/root/.cache/huggingface
-p 8000:8000 --ipc=host vllm/vllm-openai:v0.8.5 --model Qwen/Qwen3-30B-A3B --tensor-parallel-size 4 --dtype=half --enable-reasoning --reasoning-parser deepseek_r1 --max-model-len 32768 --enforce-eager --no-enable-chunked-prefill --max-model-len 16384

[houseroad]

Try --dtype float32?

[suwenzhuo]

I save tihs by using --no-enable-chunked-prefill

examlpe: vllm serve /root/model/Qwen3-8B --dtype half --port 8075 --gpu-memory-utilization 0.8 --no-enable-chunked-prefill --max-model-len 8000

The possible reason is that v100 does not support chunked-prefill