Hi, thanks for your excellent work. You use 'DataParallel' to support multi-gpu training. The standard way for saving models with 'DataParallel' is to use torch.save(model.module.state_dict(), PATH). However, in your code, you just use torch.save({ 'discriminator': self.discriminator.state_dict() }, self.dis_weights_path).
Hi, thanks for your excellent work. You use 'DataParallel' to support multi-gpu training. The standard way for saving models with 'DataParallel' is to use
torch.save(model.module.state_dict(), PATH)
. However, in your code, you just usetorch.save({ 'discriminator': self.discriminator.state_dict() }, self.dis_weights_path)
.