def test_build_graph(self, gan_type):
if tf.executing_eagerly():
# `tfgan.gan_model` doesn't work when executing eagerly.
return
hparams = self.hparams._replace(max_number_of_steps=0,
gan_type=gan_type)
train_lib.train(hparams)
def test_build_graph(self, gan_type, mock_data_provider):
hparams = self.hparams._replace(max_number_of_steps=0,
gan_type=gan_type)
# Mock input pipeline.
mock_imgs = np.zeros([hparams.batch_size, 28, 28, 1], dtype=np.float32)
mock_lbls = np.concatenate(
(np.ones([hparams.batch_size, 1], dtype=np.int32),
np.zeros([hparams.batch_size, 9], dtype=np.int32)),
axis=1)
mock_data_provider.provide_data.return_value = (mock_imgs, mock_lbls)
train_lib.train(hparams)
def test_run_one_train_step(self, mock_data_provider):
hparams = self.hparams._replace(max_number_of_steps=1,
gan_type='unconditional',
batch_size=5,
grid_size=1)
tf.compat.v1.set_random_seed(1234)
# Mock input pipeline.
mock_imgs = np.zeros([hparams.batch_size, 28, 28, 1], dtype=np.float32)
mock_lbls = np.concatenate(
(np.ones([hparams.batch_size, 1], dtype=np.int32),
np.zeros([hparams.batch_size, 9], dtype=np.int32)),
axis=1)
mock_data_provider.provide_data.return_value = (mock_imgs, mock_lbls)
train_lib.train(hparams)
def test_build_graph(self, gan_type, mock_data_provider):
if tf.executing_eagerly():
# `tfgan.gan_model` doesn't work when executing eagerly.
return
hparams = self.hparams._replace(max_number_of_steps=0,
gan_type=gan_type)
# Mock input pipeline.
mock_imgs = np.zeros([hparams.batch_size, 28, 28, 1], dtype=np.float32)
mock_lbls = np.concatenate(
(np.ones([hparams.batch_size, 1], dtype=np.int32),
np.zeros([hparams.batch_size, 9], dtype=np.int32)),
axis=1)
mock_data_provider.provide_data.return_value = (mock_imgs, mock_lbls)
train_lib.train(hparams)
def main(_):
hparams = train_lib.HParams(FLAGS.batch_size, FLAGS.train_log_dir,
FLAGS.max_number_of_steps, FLAGS.gan_type,
FLAGS.grid_size, FLAGS.noise_dims)
train_lib.train(hparams)
def test_run_one_train_step(self):
if tf.executing_eagerly():
# `tfgan.gan_model` doesn't work when executing eagerly.
return
train_lib.train(self.hparams)