运行脚本.txt

CUDA_VISIBLE_DEVICES=3 deepspeed --master_port 12345 --num_gpus=1 train.py \
    --deepspeed conf/mydeepspeed.json

###奖励模型-训练
CUDA_VISIBLE_DEVICES=0 python3 reward.py \
    --lr_scheduler_type cosine \
    --learning_rate 5e-4 \
    --do_train \
    --do_eval \
    --output_dir /home/xxm/fsdownload/chatglm2_lora/output/reward \
    --use_v2 \
    --model_path /home/xxm/model/chatglm2-6b \
    --checkpoint_dir /home/xxm/fsdownload/chatglm2_lora/output/estate_qa0 \
    --num_train_epochs 5.0 \
    --save_steps 500 \
    --reward_filename /home/xxm/fsdownload/chatglm2_lora/data/estate_reward.json \
    --per_device_train_batch_size 8 \
    --fp16

###奖励模型-推理
CUDA_VISIBLE_DEVICES=0 python3 inference_rm.py \
    --model_name_or_path /home/xxm/fsdownload/chatglm2_lora/output/merged_chatglm2_lora \
    --use_v2 \
    --reward_model_name_or_path /home/xxm/fsdownload/chatglm2_lora/output/reward


CUDA_VISIBLE_DEVICES=0 python3 merge_lora2base.py \
  --base_model_name_or_path /home/xxm/model/new/chatglm-6b \
  --peft_model_path /home/xxm/下载/ChatGLM-Efficient-Tuning/output/lora_estate_qa5 \
  --output_dir /home/xxm/fsdownload/chatglm2_lora/output/merged_chatglm_lora \
  --model_type chatglm


  wget https://developer.download.nvidia.com/compute/cuda/11.8.0/local_installers/cuda_11.8.0_520.61.05_linux.run
sudo sh cuda_11.8.0_520.61.05_linux.run