请问单机两卡（A40）训练stage one ,都需要修改哪些参数呢？

[Sarah-air]

命令行改为python -m torch.distributed.launch --nproc_per_node=2 --master_port=21 train.py -opt options/train/GoPro_S1.yml --launcher pytorch为什么报错啊？
ERROR:torch.distributed.elastic.multiprocessing.api:failed (exitcode: 1) local_rank: 0 (pid: 223240) of binary: /root/anaconda3/envs/hi_diff/bin/python

[zhengchen1999]

https://github.com/zhengchen1999/HI-Diff/blob/main/options/train/GoPro_S1.yml#L22

batch_size_per_gpu: 32 （满足总batch=64）

[Sarah-air]

https://github.com/zhengchen1999/HI-Diff/blob/main/options/train/GoPro_S1.yml#L22

batch_size_per_gpu: 32 （满足总batch=64）

好的，多谢