[The current device architecture is not supported by sycl_ext_oneapi_device_architecture]

[ammrabet]

Hello,

I am using intelanalytics/ipex-llm-serving-xpu docker image to run VLLM on the Battlemage card B580

i got this error :

terminate called after throwing an instance of 'sycl::_V1::exception'
what(): The current device architecture is not supported by sycl_ext_oneapi_device_architecture.
Aborted (core dumped)

command line :

python3 ./benchmark_vllm_throughput.py --backend vllm --dataset ./ShareGPT_V3_unfiltered_cleaned_split.json --model deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B --num-prompts 100 --seed 42 --trust-remote-code --enforce-eager --dtype float16 --device xpu

Thank you in advance
Amine mrabet

[gc-fu]

Hi, we will investigate the issue.

[gc-fu]

Hi, we have not encountered this problem after trying with our newest image: intelanalytics/ipex-llm-serving-xpu:2.2.0-b13.

Try pull this image and see if this problem persists~

The command:

python3 ./benchmark_vllm_throughput.py --backend vllm --dataset ./ShareGPT_V3_unfiltered_cleaned_split.json --model /llm/models/DeepSeek-R1-Distill-Qwen-1.5B --num-prompts 100 --seed 42 --trust-remote-code --enforce-eager --dtype float16 --device xpu --gpu-memory-utilization 0.95 --max-model-len 3000 --max-num-batched-tokens 3000