Invoking the training

All the functions we illustrated previously are created inside a DiscoGAN() class with the important parameter values declared in the __init__ function, as shown in the following code block that follows. The only two parameters that needs to be passed while training the network are the dataset_dir and the number of epochs for which the training needs to be run

   def __init__(self,dataset_dir,epochs=200):
        # Input shape
        self.dataset_dir = dataset_dir
        self.lambda_l2 = 1.0
        self.image_size = 64
        self.input_dim = 3
        self.output_dim = 3
        self.batch_size = 64 
        self.df = 64
        self.gf = 64
        self.channels = 3
        self.output_c_dim = 3
        self.l_r = 2e-4
        self.beta1 = 0.5
        self.beta2 = 0.99
        self.weight_decay = 0.00001
        self.epoch = epochs
        self.train_size = 10000
  ...