提交 ba2760dc 编写于 作者: U u010280923

update readme.md

上级 fa6038f4
...@@ -77,7 +77,7 @@ python train_rm.py --load_model "./out_sft/rwkv-190.pth" --wandb "" --proj_dir ...@@ -77,7 +77,7 @@ python train_rm.py --load_model "./out_sft/rwkv-190.pth" --wandb "" --proj_dir
### PPO Model (Reinforcement learning from Human Feedback) ### PPO Model (Reinforcement learning from Human Feedback)
``` ```
python train_rm.py --load_sft_model "./out_sft/rwkv-190.pth" --load_rm_model "./out_rm/rm-2.pth" --wandb "" \ python train_ppo.py --load_sft_model "./out_sft/rwkv-190.pth" --load_rm_model "./out_rm/rm-2.pth" --wandb "" \
--proj_dir "out_rlhf" \ --proj_dir "out_rlhf" \
--data_file "data/rm_mock_data.csv" --data_type "utf-8" --vocab_size 50277 \ --data_file "data/rm_mock_data.csv" --data_type "utf-8" --vocab_size 50277 \
--ctx_len 1024 --epoch_steps 200 --epoch_count 1000 --epoch_begin 0 --epoch_save 2 \ --ctx_len 1024 --epoch_steps 200 --epoch_count 1000 --epoch_begin 0 --epoch_save 2 \
......
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册