Discriminator

With the generator defined, the next step in building a GAN is to construct the discriminator. It takes the generator's output as input, and produces a binary prediction: is the input generated or real?

You will find torch.nn imported already imported for you as nn. You can also access a custom disc_block() function which returns a block of a linear layer followed by a LeakyReLU activation. You will use it as a building block for the discriminator.

def disc_block(in_dim, out_dim):
    return nn.Sequential(
        nn.Linear(in_dim, out_dim),
        nn.LeakyReLU(0.2)
    )

This exercise is part of the course

Deep Learning for Images with PyTorch

View Course

Exercise instructions

Add the last discriminator block to the model, with the appropriate input size and the output of 256.
After the last discriminator block, add a linear layer to map the output to the size of 1.
Define the forward() method to pass the input image through the sequential block defined in __init__().

Hands-on interactive exercise

Have a go at this exercise by completing this sample code.

class Discriminator(nn.Module):
    def __init__(self, im_dim):
        super(Discriminator, self).__init__()
        self.disc = nn.Sequential(
            disc_block(im_dim, 1024),
            disc_block(1024, 512),
            # Define last discriminator block
            ____,
            # Add a linear layer
            ____,
        )

    def forward(self, x):
        # Define the forward method
        ____

Edit and Run Code