MNIST-Keras-GAN-Model.py

LATENT_SIZE = 16

adam = keras.optimizers.Adam(lr=0.0002, beta_1=0.5)

with tf.device("/gpu:0"):
    input_x_img = keras.layers.Input((784,))
    x           = keras.layers.Dense(1024, activation="relu")(input_x_img)
    x           = keras.layers.Dense(256, activation="relu")(x)
    x           = keras.layers.Dense(1, activation="sigmoid")(x)
    discriminator = keras.models.Model(inputs=input_x_img, outputs=x)
    discriminator.compile(optimizer=adam, loss="binary_crossentropy")

    input_x_latent = keras.layers.Input((LATENT_SIZE,))
    x              = keras.layers.Dense(256, activation="relu")(input_x_latent)
    x              = keras.layers.Dense(1024, activation="relu")(x)
    x              = keras.layers.Dense(784, activation="tanh")(x)
    generator = keras.models.Model(inputs=input_x_latent, outputs=x)
    generator.compile(optimizer=adam, loss="binary_crossentropy")

    discriminator.trainable = False
    gan_input  = keras.layers.Input((LATENT_SIZE,))
    x          = generator(gan_input)
    gan_output = discriminator(x)
    gan = keras.models.Model(inputs=gan_input, outputs=gan_output)
    gan.compile(optimizer=adam, loss="binary_crossentropy")