def train_cifar_gan(args):
cifar_ds = CifarDataset(args.cifar_ds_path, 64, args.how_many_labeled)
learning_options = adam_learning_options()
learning_options['lr'] = args.learning_rate
gan_cifar_trainer = SSGanTrainer(args.latent_dim, cifar_ds, 10,
args.out_weights_dir, 64,
learning_options)
gan_cifar_trainer.train_ss(args.epochs_num)
def train_cifar_gan(args):
cifar_ds = SupervisedCifarDataset(args.cifar_ds_path, 64,
args.how_many_labeled)
learning_options = adam_learning_options()
learning_options['lr'] = args.learning_rate
cifar_trainer = SupCnnTrainer(cifar_ds, 64, cifar_gan_discriminator,
learning_options)
cifar_trainer.train(args.epochs_num, args.testing_interval)
def __init__(self,
dataset,
batch_size,
create_model_func,
train_options=adam_learning_options()):
self.dataset = dataset
self.batch_size = batch_size
self.train_options = train_options
self.img_size = dataset.img_size()
self.classes_num = dataset.classes_num()
self.input_images = tf.placeholder(
tf.float32,
shape=[batch_size, self.img_size, self.img_size, 3],
name="input_images")
self.labels = tf.placeholder(tf.float32,
shape=[batch_size, self.classes_num])
self.create_model_func = create_model_func
self.logger.info("Initialization finished, dataset size = %d" %
dataset.size())
def __init__(self,
latent_dim,
dataset,
classes_num,
out_weights_dir,
batch_size=64,
train_options=adam_learning_options()):
super(SSGanTrainer,
self).__init__(latent_dim, dataset, out_weights_dir, batch_size,
train_options)
self.dataset = dataset
self.classes_num = classes_num
self.testing_interval = config.TESTING_INTERVAL
self.testing_iterations = config.TESTING_ITERATIONS
self.labels = tf.placeholder(dtype=tf.float32,
shape=[self.batch_size, self.classes_num])
self.unlabeled_images = tf.placeholder(
dtype=tf.float32,
shape=[self.batch_size, self.img_size, self.img_size, 3],
name='unlabeled_images')
def __init__(self,
latent_dim,
dataset,
out_weights_dir,
train_options=adam_learning_options(),
kl_loss_weight=0.0005):
self.out_weights_dir = out_weights_dir
self.train_options = train_options
self.latent_dim = latent_dim
self.batch_size = dataset.batch_size
self.logger.info("Creating output weights directory %s" %
self.out_weights_dir)
tl.files.exists_or_mkdir(out_weights_dir)
self.img_size = dataset.img_size()
self.input_images = tf.placeholder(
tf.float32,
shape=[self.batch_size, self.img_size, self.img_size, 3],
name='input_images')
self.kl_loss_weight = kl_loss_weight
self.weights_dump_interval = config.WEIGHTS_DUMP_INTERVAL
self.dataset = dataset
def __init__(self, latent_dim, transformer, out_weights_dir, batch_size=64, train_options=adam_learning_options()):
self.out_weights_dir = out_weights_dir
self.transformer = transformer
self.latent_dim = latent_dim
self.batch_size = batch_size
self.logger.info("Creating output weights directory %s" % self.out_weights_dir)
tl.files.exists_or_mkdir(out_weights_dir)
self.img_size = transformer.img_size()
self.train_options = train_options
self.weights_dump_interval = config.WEIGHTS_DUMP_INTERVAL
self.input_images = tf.placeholder(tf.float32, shape=[self.batch_size, self.img_size, self.img_size, 3],
name='input_images')
self.z = tf.placeholder(tf.float32, shape=[None, self.latent_dim], name='z')
def train_cifar_gan(args):
cifar_ds = UnsupervisedCifarDataSet(args.cifar_ds_path, 64)
learning_options = adam_learning_options()
learning_options['lr'] = args.learning_rate
vae_cifar_trainer = VaeTrainer(args.latent_dim, cifar_ds, args.out_weights_dir, learning_options)
vae_cifar_trainer.train(args.epochs_num)