First, we feed the noise, and the conditional variable, , to the generator, and it will output the fake image, that is, :
fake_x = generator(z, c)
Now we feed the real image along with conditional variable, , to the discriminator, , and get the probability of them being real:
D_logits_real = discriminator(x,c)
Similarly, we feed the fake image, fake_x, and the conditional variable, , to the discriminator, , and get the probability of them being real:
D_logits_fake = discriminator(fake_x, c, reuse=True)