model_stats['d_train_loss'].append(d_loss)
model_stats['d_train_acc'].append(d_acc)
# set the discriminator to not trainable
discriminator.trainable = False
# discriminator training
g_noise = np.random.normal(0, 1, (batch_size, latent_dim))
# g_valid = [1, 0]
g_valid = np.concatenate((np.ones(
(batch_size, 1)), np.zeros((batch_size, 1))),
axis=1)
# train the generator
g_loss, g_acc = generator_discriminator.train_on_batch(g_noise, g_valid)
model_stats['g_train_loss'].append(g_loss)
model_stats['g_train_acc'].append(g_acc)
if epoch_idx % gen_epoch == 0 and epoch_idx > 0:
plot_generated_img_samples(None,
generator.predict(gen_noise).reshape(
(-1, img_rows, img_cols)),
to_save=True,
iteration=epoch_idx,
model_name=model_name)
if verbose:
print('{}Epoch {} Discriminator Loss: {:2.4f}, Acc: {:2.4f}.'.format(
print_pad(1), print_epoch, d_loss, d_acc))
# set the discriminator to not trainable
discriminator.trainable = False
# discriminator training
g_noise = np.random.normal(0, 1, (batch_size, latent_dim))
# g_valid = [1, 0]
g_valid = np.concatenate((np.ones(
(batch_size, 1)), np.zeros((batch_size, 1))),
axis=1)
# train the generator
g_loss, g_acc = generator_discriminator.train_on_batch(
g_noise,
g_valid,
epoch_num=epoch_idx_p1,
batch_num=epoch_idx_p1,
batch_size=batch_size)
model_stats['g_train_loss'].append(g_loss)
model_stats['g_train_acc'].append(g_acc)
if epoch_idx % gen_epoch == 0 and epoch_idx > 0:
plot_generated_img_samples(None,
generator.predict(gen_noise).reshape(
(-1, img_rows, img_cols)),
to_save=True,
iteration=epoch_idx,
model_name=model_name)
if verbose:
