模型输出无意义的内容

[altqxd]

从huggingface现在alpaca的模型之后 按照torchrun --nproc-per-node=1 demos/single_turn.py --llama_config /root/autodl-tmp/model/LLaMA2-Accessory/config/7B_params.json --tokenizer_path /root/autodl-tmp/model/LLaMA2-Accessory/config/tokenizer.model --pretrained_path /root/autodl-tmp/model/LLaMA2-Accessory/finetune/sg/alpaca 官网提供的命令运行，可以正常启动gradio，但是模型输出的是无意义的内容

[ChrisLiu6]

Hi, note that we release delta patches of our trained checkpoints instead of off-the-shelf checkpoints to comply with llama's license. Please follow the instructions here to merge the weights before inference.

[altqxd]

Hi, note that we release delta patches of our trained checkpoints instead of off-the-shelf checkpoints to comply with llama's license. Please follow the instructions here to merge the weights before inference.

thanks,i will try