D( G(z).detach() ), added D means for monitoring

and use os.makedirs py2.7 compatible

Author: tomsercu

dcgan/main.py

@@ -35,7 +35,10 @@
 opt = parser.parse_args()
 print(opt)
 
-os.makedirs(opt.outf, exist_ok=True)
+try:
+ os.makedirs(opt.outf)
+except OSError:
+ pass
 opt.manualSeed = random.randint(1, 10000) # fix seed
 print("Random Seed: ", opt.manualSeed)
 random.seed(opt.manualSeed)
@@ -205,16 +208,17 @@ def forward(self, input):
  output = netD(input)
  errD_real = criterion(output, label)
  errD_real.backward()
+ D_x = output.data.mean()
 
  # train with fake
  noise.data.resize_(batch_size, nz, 1, 1)
  noise.data.normal_(0, 1)
- fake = netG(noise)
- input.data.copy_(fake.data)
+ fake = netG(noise).detach()
  label.data.fill_(fake_label)
- output = netD(input)
+ output = netD(fake)
  errD_fake = criterion(output, label)
  errD_fake.backward()
+ D_G_z1 = output.data.mean()
  errD = errD_real + errD_fake
  optimizerD.step()
 
@@ -228,11 +232,12 @@ def forward(self, input):
  output = netD(fake)
  errG = criterion(output, label)
  errG.backward()
+ D_G_z2 = output.data.mean()
  optimizerG.step()
 
- print('[%d/%d][%d/%d] Loss_D: %f Loss_G: %f'
+ print('[%d/%d][%d/%d] Loss_D: %.4f Loss_G: %.4f D(x): %.4f D(G(z)): %.4f / %.4f'
  % (epoch, opt.niter, i, len(dataloader),
- errD.data[0], errG.data[0]))
+ errD.data[0], errG.data[0], D_x, D_G_z1, D_G_z2))
  if i % 100 == 0:
  vutils.save_image(real_cpu,
  '%s/real_samples.png' % opt.outf)