Dis.conv_pool_block(n_filter=8, filter_size=(7,7))
Dis.fully_connected(4)
Dis.fully_connected(1) # for W-Loss
Dis.build()
Gen = Painter(Dis, input_shape, output_shape)
Gen.input()
Gen.conv_block(16, filter_size=(5,5), padding="same")
Gen.conv_block(16, filter_size=(7,7), padding="same")
Gen.pooling_block((2,2))
Gen.conv_block(32, filter_size=(9,9), padding="same")
Gen.conv_block(32, filter_size=(11,11), padding="same")
Gen.pooling_block((2,2))
Gen.top_block()
#Gen.build_monitored([2,5],[0.5,0.5])
Gen.build()
painter_optimizer = Adam(learning_rate=0.001, beta_1=0.9, beta_2=0.999, epsilon=1e-07)
disc_optimizer = Adam(learning_rate=0.001, beta_1=0.9, beta_2=0.999, epsilon=1e-07)
for i in range(10):
print("Training Discriminator...")
for dis_training_loop in range(25):
discri_grads = Dis.get_gradient()
disc_optimizer.apply_gradients(zip(discri_grads, Dis.model.trainable_weights))
print('\tDiscriminator loss: ',Dis.__current_loss__)
print("Training Generator...")
