how to solve this problem?when I run main.sh

[Sunmingy]

[mathildecaron31]

Hi @Sunmingy,
If you're not launching main.sh to a SLURM cluster but locally then you should use torch.distributed.launch. For example replace

DeeperCluster/main.sh

Line 16 in d38ada1

python main.py

by python -m torch.distributed.launch --nproc_per_node=$NGPU main.py with $NGPU being the number of gpus you have.