Hands-On Generative Adversarial Networks with Keras

Training

Our training method consists of the following two parts:

• Initializing the required variables and compiling models
• The training loop itself

We will start by looking at the training layout:

def train(n_channels=3, resolution=32, z_dim=128, n_labels=0, lr=1e-3,          e_drift=1e-3, wgp_target=750, initial_resolution=4, total_kimg=20000,          training_kimg=400, transition_kimg=400, iters_per_checkpoint=500,          n_checkpoint_images=16, glob_str='cifar10', out_dir='cifar10'):

Let's examine each argument of our train method in the following table:

Argument	Description
n_channels	The number of image channels
resolution	The target image resolution
z_dim	The size of the latent vector
n_labels	The number of labels
lr	The learning rate for the ...