MCPcopy Create free account
hub / github.com/JunlinHan/DCLGAN / backward_D_basic

Method backward_D_basic

models/simdcl_model.py:189–208  ·  view source on GitHub ↗

Calculate GAN loss for the discriminator Parameters: netD (network) -- the discriminator D real (tensor array) -- real images fake (tensor array) -- images generated by a generator Return the discriminator loss. We also call loss_D.ba

(self, netD, real, fake)

Source from the content-addressed store, hash-verified

187 self.idt_B = self.netG_B(self.real_A)
188
189 def backward_D_basic(self, netD, real, fake):
190 """Calculate GAN loss for the discriminator
191 Parameters:
192 netD (network) -- the discriminator D
193 real (tensor array) -- real images
194 fake (tensor array) -- images generated by a generator
195
196 Return the discriminator loss.
197 We also call loss_D.backward() to calculate the gradients.
198 """
199 # Real
200 pred_real = netD(real)
201 loss_D_real = self.criterionGAN(pred_real, True)
202 # Fake
203 pred_fake = netD(fake.detach())
204 loss_D_fake = self.criterionGAN(pred_fake, False)
205 # Combined loss and calculate gradients
206 loss_D = (loss_D_real + loss_D_fake) * 0.5
207 loss_D.backward()
208 return loss_D
209
210 def backward_D_A(self):
211 """Calculate GAN loss for discriminator D_A"""

Callers 2

backward_D_AMethod · 0.95
backward_D_BMethod · 0.95

Calls 1

backwardMethod · 0.80

Tested by

no test coverage detected