search

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
3.21k stars 488 forks source link

训练奖励模型的时候，如果训练百川模型，请问model_type 传什么参数？我传的 baichuan 这个报错了。请问怎么处理？ #144

Closed DaiJitao closed 1 year ago

DaiJitao commented 1 year ago

Describe the bug

Please provide a clear and concise description of what the bug is. If applicable, add screenshots to help explain your problem, especially for visualization related problems.

DaiJitao commented 1 year ago

我这样传参数的： CUDA_VISIBLE_DEVICES=$gpu python -u reward_modeling.py \ --model_type baichuan \ --model_name_or_path merged-sft \ --train_file_dir ./data/reward \ --validation_file_dir ./data/reward \ --per_device_train_batch_size 4 \ --per_device_eval_batch_size 4 \ --do_train \ --use_peft True \ --seed 42 \ --max_train_samples 1000 \ ...

DaiJitao commented 1 year ago

--model_type 传什么参数？

shibing624 commented 1 year ago

百川跟chatglm一样，暂不支持reward training。 https://github.com/shibing624/MedicalGPT/wiki/FAQ