Closed TY-hue123 closed 1 year ago
I tried to use the pre-trained model (https://huggingface.co/Chinese-Vicuna/Chinese-Vicuna-lora-7b-belle-and-guanaco) to fine-tune on my own dataset but it cannot learn a good talking behavior:
@TY-hue123 Maybe it's a problem with your network? You can try downloading the model file and loading it offline。
@SHITIANYU-hue You can refer to this issue(https://github.com/Facico/Chinese-Vicuna/issues/121#issuecomment-1534012359) Check your GPU and tokenizer. You can try using pure cpu inference to see whether you can produce normal results
i tried to change the model's path as MODEL_PATH="Facico/Chinese-Vicuna-lora-7b-3epoch-belle-and-guanaco", but it can't find the model:
https://github.com/Facico/Chinese-Vicuna/blob/master/finetune.sh#L8