launch.py: error: argument --nproc_per_node: invalid int value: ''

[Inuk Kang]

I am running training data. but got this error. do you know how to deal with?

[cfzd]

@InukKang
Hi, how many GPUs do you have? You can also try to set the number of GPUs manually:

python -m torch.distributed.launch --nproc_per_node=4 main.py -a fcanet50 --dali_cpu --b 64 /path/to/your/ImageNet

The --nproc_per_node=4 set nproc_per_node to 4.

[Inuk Kang]

Thank you!