Text Generation
Transformers
PyTorch
Chinese
English
llama
text-generation-inference

继续微调的问题

#9
by yuqin - opened

请问,我可以基于这个模型再用我自己的数据进行微调吗

能给一份处理之后的,微调数据格式样例吗?

数据格式是一个用json保存的列表。列表中的每项是一个openai api格式保存的对话,格式如下:

dummy_message = [{"role": "user", "content": "Who are you?"},
                 {"role": "assistant", "content": "I am vicuna, a language model trained by researchers from open-source community."},
                 {"role": "user", "content": "What can you do?"},
                 {"role": "assistant", "content": "I can chat with you."}]

Refer to https://huggingface.co/fireballoon/baichuan-vicuna-7b/blob/d3b83cf408270757cc8b6e6335d7feb45f868164/train_vicuna.py#L82

Sign up or log in to comment