[Bug][V1]: Qwen/Qwen2.5-7B-Instruct accuracy ceval-valid failed

[Yikun]

Your current environment

vLLM Ascend main, v0.9.0

🐛 Describe the bug

https://github.com/vllm-project/vllm-ascend/actions/runs/15397909461/attempts/2#summary-43344496525

🎯 Qwen2.5-7B-Instruct Accuracy Test

vLLM version: vLLM: 0.9.0, vLLM Ascend: refs/pull/1040/merge

Software Environment: CANN: 8.1.RC1, PyTorch: 2.5.1, torch-npu: 2.5.1

Hardware Environment: Atlas A2 Series

Datasets: ceval-valid,gsm8k

Command:

export MODEL_ARGS='pretrained=Qwen/Qwen2.5-7B-Instruct, max_model_len=4096,dtype=auto,tensor_parallel_size=2,gpu_memory_utilization=0.6'
lm_eval --model vllm --modlel_args $MODEL_ARGS --tasks ceval-valid,gsm8k \ 
--apply_chat_template --fewshot_as_multiturn --num_fewshot 5 --batch_size 1

 

Task	Filter	n-shot	Metric	Value	Stderr
ceval-valid(V1)	none	5	acc	0.2303 <-----	± 0.0115
ceval-valid(V0)	none	5	acc	0.8001	± 0.0105