torchrun --standalone --nproc_per_node=2 train_gpt2.py

torchrun --standalone --nproc_per_node=2 train_gpt2_simple.py

torchrun --standalone --nproc_per_node=2 train_CampGPT_X.py

python train_CampGPT_X.py


CUDA_VISIBLE_DEVICES=1 python train_CampGPT_X_plus.py
CUDA_VISIBLE_DEVICES=0,1 torchrun --nproc_per_node=2 train_CampGPT_X_plus.py

CUDA_VISIBLE_DEVICES=0,1,2,3 torchrun --nproc_per_node=4 train_CampGPT_X_plus.py


# 按顺序执行
CUDA_VISIBLE_DEVICES=0,1,2,3 torchrun --nproc_per_node=4 train_CampGPT_X_plus.py
CUDA_VISIBLE_DEVICES=1 python train_sft.py
CUDA_VISIBLE_DEVICES=1 python train_dpo.py
CUDA_VISIBLE_DEVICES=1 python export_hf.py
# cli single api
CUDA_VISIBLE_DEVICES=1 python serve.py --mode cli
CUDA_VISIBLE_DEVICES=1 python serve.py --mode single
CUDA_VISIBLE_DEVICES=1 python serve.py --mode api
CUDA_VISIBLE_DEVICES=1 python test_sft.py
CUDA_VISIBLE_DEVICES=1 python test_dpo.py

bash campgpt-student-handbook/upload.sh