自定义评测数据集做评测时出现，模型用vllm.entrypoints.openai.api_server起的。运行评测脚本出现错误

[Wolverhampton0]

模型启动方法：python -m vllm.entrypoints.openai.api_server --served-model-name qwen2-7b-instruct --model /app/Qwen2-7B-Instruct --gpu-memory-utilization 0.9
评测方法：swift eval --eval_url http://127.0.0.1:8000/v1 --eval_is_chat_model true --model_type qwen2-7b-instruct --eval_dataset no --custom_eval_config /app/swift/examples/pytorch/llm/eval_example/custom_config.json
配置文件：[
{
"name": "custom_general_qa",
"pattern": "general_qa",
"dataset": "/app/swift/examples/pytorch/llm/eval_example/custom_general_qa",
"subset_list": ["default"]
},
{
"name": "custom_ceval",
"pattern": "ceval",
"dataset": "/app/swift/examples/pytorch/llm/eval_example/custom_ceval",
"subset_list": ["default"]
}
]

报错：future: <Task finished name='Task-26' coro=<EvalModel._call_openai() done, defined at /app/swift/swift/llm/eval.py:44> exception=HTTPError("[{'type': 'extra_forbidden', 'loc': ('body', 'num_beams'), 'msg': 'Extra inputs are not permitted', 'input': 1}]")>
Traceback (most recent call last):
File "/app/swift/swift/llm/eval.py", line 48, in _call_openai
resp = await inference_client_async(
File "/app/swift/swift/llm/utils/client_utils.py", line 309, in inference_client_async
raise HTTPError(resp_obj['message'])
requests.exceptions.HTTPError: [{'type': 'extra_forbidden', 'loc': ('body', 'num_beams'), 'msg': 'Extra inputs are not permitted', 'input': 1}]

[Wolverhampton0]

同样的模型启动方式，评测使用ARC-e就没有问题。
swift eval --eval_url http://127.0.0.1:8000/v1 --eval_is_chat_model true --model_type qwen2-7b-instruct --eval_dataset ARC_e