Why we do not need to use DistributedDataParallel for netF?

[rakhimovv]

Thanks for the great project. Looking through code, I noticed that in SRGAN_model.py there are such lines:

if opt['dist']:
    pass  # do not need to use DistributedDataParallel for netF
else:
    self.netF = DataParallel(self.netF)

can you please provide motivation of not wrapping netF with DistributedDataParallel

[xinntao]

Because there is no need to update netF during training.
If we use DistributedDataParallel for netF and we do not update netF, then errors will be raised.

[rakhimovv]

Thanks for the answer!