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。
3.38k
stars
508
forks
source link
RuntimeError: "nll_loss_out_frame" not implemented for 'Half' #387
Open
Li-Jicheng opened 5 months ago
跑https://colab.research.google.com/github/shibing624/MedicalGPT/blob/main/run_training_dpo_pipeline.ipynb#scrollTo=J5kYehpzESyt (run_training_dpo_pipeline.ipynb)这个脚本时,pretrain阶段报错