hello @AaronYALai , in run_InfoGAN.py, 'D_loss=D_criterion(outputs[:,0],labels)' is the mean of 2batchSize loss between outputs and labels, but I think D_loss=truth_input_loss+fake_input_loss, that should be the mean of 1batchSize loss. I'm looking forward to your reply.
hello @AaronYALai , in run_InfoGAN.py, 'D_loss=D_criterion(outputs[:,0],labels)' is the mean of 2batchSize loss between outputs and labels, but I think D_loss=truth_input_loss+fake_input_loss, that should be the mean of 1batchSize loss. I'm looking forward to your reply.