1. 相同目录下文件 无法引用
设定当前目录为跟目录
2. 添加新的环境变量
3. Can’t run remote python interpreter: Can’t get remote credentials for deployment server Copy of xxx
将project interpreter 中不用的解释器删掉
4. pycharm 带参数调试
(1)打开edict config
(2)输入参数形式
--stage ppo --model_name_or_path baichuan/Baichuan2-13B-Chat --do_train --dataset jt_rl --dataset_dir llm_finetune/LLaMA-Efficient-Tuning-main/data --template baichuan --finetuning_type lora --lora_target W_pack,o_proj,gate_proj,up_proj,down_proj --resume_lora_training False --checkpoint_dir llm_finetune/LLaMA-Efficient-Tuning-main/output/baichuan_lora_mix_1025 --reward_model LLaMA-Efficient-Tuning-main/output/reward_1025 --output_dir llm_finetune/LLaMA-Efficient-Tuning-main/output/ppo --per_device_train_batch_size 2 --gradient_accumulation_steps 4 --lr_scheduler_type cosine --logging_steps 10 --save_steps 1000 --learning_rate 1e-5 --num_train_epochs 1.0 --plot_loss --fp16