First, we feed the noise to the generator and it will output the fake image, :

fake_x = generator(z, z_dim, batch_size, is_training=is_training)

Now we feed the real image to the discriminator and get the probability of the real image being real:

D_logit_real = discriminator(x, reuse=False, is_training=is_training)

Similarly, we feed the fake image to the discriminator, , and get the probability of the fake image being real:

D_logit_fake = discriminator(fake_x, reuse=True,  is_training=is_training)