diff --git a/configs/server_config.py.example b/configs/server_config.py.example index 714a343..6f5fb0d 100644 --- a/configs/server_config.py.example +++ b/configs/server_config.py.example @@ -40,6 +40,8 @@ FSCHAT_MODEL_WORKERS = { "device": LLM_DEVICE, # False,'vllm',使用的推理加速框架,使用vllm如果出现HuggingFace通信问题,参见doc/FAQ # vllm对一些模型支持还不成熟,暂时默认关闭 + # fschat=0.2.33的代码有bug, 如需使用,源码修改fastchat.server.vllm_worker, + # 将103行中sampling_params = SamplingParams的参数stop=list(stop)修改为stop= [i for i in stop if i!=""] "infer_turbo": False, # model_worker多卡加载需要配置的参数