pytorch-StarGAN-VC
pytorch-StarGAN-VC copied to clipboard
The classification loss in generator
According to the original paper,
line 206 and 207 in solver.py should be changed into following:
out_cls = self.C(x_fake)
g_loss_cls = CELoss(input=out_cls, target=speaker_idx_trg)
so that the g_loss_cls would train the generator.