shibing624 / MedicalGPT

MedicalGPT: Training Your Own Medical GPT Model with ChatGPT Training Pipeline. 训练医疗大模型,实现了包括增量预训练(PT)、有监督微调(SFT)、RLHF、DPO、ORPO。
Apache License 2.0
2.94k stars 452 forks source link

这是为什么啊?!用指令SFT后,模型的回答都变得短了,短就算了,还不对, 新的内容没学习进去 #306

Closed Thunderltx closed 5 months ago

Thunderltx commented 5 months ago

Describe the Question

Please provide a clear and concise description of what the question is. 数据集是shibing624/medical/train_zh_0.json 模型用的是baichuan 这是SFT后 image 这是SFT前 image (1)

Thunderltx commented 5 months ago

再补一句,不是全参,是LoRA

shibing624 commented 5 months ago

看下wiki,加些通用数据。多训练几轮(eg:epochs=10)

Thunderltx commented 5 months ago

看下wiki,加些通用数据。多训练几轮(eg:epochs=10)

谢谢答复,不过SFT也要加通用的数据么

shibing624 commented 5 months ago