Can multiple GPUs be utilized with the PyTorch DataParallel class?

[Tpool1]

Hello,

I am looking to use the 3D_GMIC model with a batch size greater than 1 by using the PyTorch DataParallel class and multiple GPUs. I noticed there are assertions at the beginning of the forward function expecting a batch size of only 1. If a higher batch size has been successfully employed, assistance would be greatly appreciated!