Why move the tensor to cuda:0?

[Kelfvin]

In many files you specify cuda:0, why do you need to do this?

I've found during training that this causes HiFiNet and the weights of the IsolatingLossFunction in it to not be on the same device. I hope to get your reply, thanks!

HiFi_IFDL/utils/custom_loss.py

Lines 11 to 12 in cf59735

	device = torch.device('cuda:0')
	device_ids = [0]