Helper methods make things in a class more efficient or frequently used throughout the method. In our case, it seemed important to ensure that we could check the structure of the model as a text output and graphically:

1. The text summary available from Keras is easy to implement:

def summary(self):
    return self.Generator.summary()

2. The summary function should return an output in the Terminal like this once we use the Generator class:

_________________________________________________________________
Layer (type) Output Shape Param # 
=================================================================
dense_1 (Dense) (None, 128) 12928 
_________________________________________________________________
leaky_re_lu_1 (LeakyReLU) (None, 128) 0 
_________________________________________________________________
batch_normalization_1 (Batch (None, 128) 512 
_________________________________________________________________
dense_2 (Dense) (None, 256) 33024 
_________________________________________________________________
leaky_re_lu_2 (LeakyReLU) (None, 256) 0 
_________________________________________________________________
batch_normalization_2 (Batch (None, 256) 1024 
_________________________________________________________________
dense_3 (Dense) (None, 512) 131584 
_________________________________________________________________
leaky_re_lu_3 (LeakyReLU) (None, 512) 0 
_________________________________________________________________
batch_normalization_3 (Batch (None, 512) 2048 
_________________________________________________________________
dense_4 (Dense) (None, 1024) 525312 
_________________________________________________________________
leaky_re_lu_4 (LeakyReLU) (None, 1024) 0 
_________________________________________________________________
batch_normalization_4 (Batch (None, 1024) 4096 
_________________________________________________________________
dense_5 (Dense) (None, 784) 803600 
_________________________________________________________________
reshape_1 (Reshape) (None, 28, 28, 1) 0 
=================================================================
Total params: 1,514,128
Trainable params: 1,510,288
Non-trainable params: 3,840
_________________________________________________________________

3. Next, let's go over the generator's version of the model saver. This function is identical to the function inside of the Discriminator, except for the path:

def save_model(self):
        plot_model(self.Discriminator.model, 
        to_file='/data/Discriminator_Model.png')

This function will output a PNG in the data folder that represents the structure of the model:

So, this was how we built the GAN generator. I hope you found it easy. We are now ready to make a masterpiece.