vanilla gan generator

vanilla_gan_generator.py

class Generator(nn.Module):
    def __init__(self, latent_dim, output_activation=None):
        """A generator for mapping a latent space to a sample space.

        Args:
            latent_dim (int): latent dimension ("noise vector")
            layers (List[int]): A list of layer widths including output width
            output_activation: torch activation function or None
        """
        super(Generator, self).__init__()
        self.linear1 = nn.Linear(latent_dim, 64)
        self.leaky_relu = nn.LeakyReLU()
        self.linear2 = nn.Linear(64, 32)
        self.linear3 = nn.Linear(32, 1)
        self.output_activation = output_activation

    def forward(self, input_tensor):
        """Forward pass; map latent vectors to samples."""
        intermediate = self.linear1(input_tensor)
        intermediate = self.leaky_relu(intermediate)
        intermediate = self.linear2(intermediate)
        intermediate = self.leaky_relu(intermediate)
        intermediate = self.linear3(intermediate)
        if self.output_activation is not None:
            intermediate = self.output_activation(intermediate)
        return intermediate