G_sample = netG.forward(noise, label)
D_real = netD.forward(mnist_img, label)
D_fake = netD.forward(G_sample.detach(), label)
D_loss_real = criterion(D_real, ones_label)
D_loss_fake = criterion(D_fake, zeros_label)
D_loss = D_loss_real + D_loss_fake
netD.zero_grad()
D_loss.backward()
D_solver.step()
netG.zero_grad()
netD.zero_grad()
G_sample = netG.forward(noise, label)
output = netD.forward(G_sample, label)
G_loss = criterion(output, ones_label)
G_loss.backward()
G_solver.step()
error info:
who can tell me why ? i have been trying to solve the problem for long. thanks very much !