# for the SSGAN we need to feed the labels, L, when initializing
gan = GAN(
'mini_imagenet',
image_shape=shape,
latent_dim=latent_dim,
database=mini_imagenet_database,
parser=mini_imagenet_parser,
generator=mini_imagenet_generator,
discriminator=mini_imagenet_discriminator,
visualization_freq=1,
d_learning_rate=0.0003,
g_learning_rate=0.0003,
)
gan.perform_training(epochs=GAN_EPOCHS, checkpoint_freq=50)
gan.load_latest_checkpoint()
print("training GAN is done")
time.sleep(1)
# Split labeled and not labeled
train_folders = mini_imagenet_database.train_folders
keys = list(train_folders.keys())
labeled_keys = np.random.choice(keys, int(len(train_folders.keys())*labeled_percentage), replace=False)
train_folders_labeled = {k: v for (k, v) in train_folders.items() if k in labeled_keys}
train_folders_unlabeled = {k: v for (k, v) in train_folders.items() if k not in labeled_keys}
mini_imagenet_database.train_folders = train_folders_labeled
L = None
ssml_maml = SSMLMAML(
omniglot_parser = OmniglotParser(shape=shape)
gan = GAN(
'omniglot',
image_shape=shape,
latent_dim=latent_dim,
database=omniglot_database,
parser=omniglot_parser,
generator=omniglot_generator,
discriminator=omniglot_discriminator,
visualization_freq=50,
d_learning_rate=0.0003,
g_learning_rate=0.0003,
)
# gan.perform_training(epochs=49, checkpoint_freq=1)
gan.load_latest_checkpoint(epoch_to_load_from='500')
maml_gan = MAMLGAN(gan=gan,
latent_dim=latent_dim,
generated_image_shape=shape,
database=omniglot_database,
network_cls=SimpleModel,
n=5,
k_ml=1,
k_val_ml=5,
k_val=1,
k_val_val=15,
k_val_test=15,
k_test=1,
meta_batch_size=4,
num_steps_ml=5,