Wasserstein_D seems have no impact when training Dicriminator？ about pytorch-wgan HOT 1 CLOSED

# Create total loss and optimize
  self.D_opt.zero_grad()
  d_loss = d_generated.mean() - d_real.mean() + gradient_penalty
  d_loss.backward()

    # Train discriminator
    # WGAN - Training discriminator more iterations than generator
    # Train with real images
    d_loss_real = self.D(images)
    d_loss_real = d_loss_real.mean(0).view(1)
    
    # Train with fake images
    z = self.get_torch_variable(torch.randn(self.batch_size, 100, 1, 1))
    fake_images = self.G(z)
    d_loss_fake = self.D(fake_images)
    d_loss_fake = d_loss_fake.mean(0).view(1)

    # Train with gradient penalty
    gradient_penalty = self.calculate_gradient_penalty(images.data, fake_images.data)

    d_loss = d_loss_fake - d_loss_real + gradient_penalty
    d_loss.backward()
    self.d_optimizer.step()