def get_dummy_gan_model():
"""Similar to get_gan_model()."""
# TODO(joelshor): Find a better way of creating a variable scope.
with tf.compat.v1.variable_scope('generator') as gen_scope:
gen_var = tf.compat.v1.get_variable('dummy_var', initializer=0.0)
with tf.compat.v1.variable_scope('discriminator') as dis_scope:
dis_var = tf.compat.v1.get_variable('dummy_var', initializer=0.0)
return tfgan.StarGANModel(
input_data=tf.ones([1, 2, 2, 3]),
input_data_domain_label=tf.ones([1, 2]),
generated_data=tf.ones([1, 2, 2, 3]),
generated_data_domain_target=tf.ones([1, 2]),
reconstructed_data=tf.ones([1, 2, 2, 3]),
discriminator_input_data_source_predication=tf.ones([1]) * dis_var,
discriminator_generated_data_source_predication=tf.ones([1]) *
gen_var * dis_var,
discriminator_input_data_domain_predication=tf.ones([1, 2]) * dis_var,
discriminator_generated_data_domain_predication=tf.ones([1, 2]) *
gen_var * dis_var,
generator_variables=[gen_var],
generator_scope=gen_scope,
generator_fn=None,
discriminator_variables=[dis_var],
discriminator_scope=dis_scope,
discriminator_fn=None)
def setUp(self):
super(StarGANLossWrapperTest, self).setUp()
self.input_data = tf.ones([1, 2, 2, 3])
self.input_data_domain_label = tf.constant([[0, 1]])
self.generated_data = tf.ones([1, 2, 2, 3])
self.discriminator_input_data_source_predication = tf.ones([1])
self.discriminator_generated_data_source_predication = tf.ones([1])
def _discriminator_fn(inputs, num_domains):
"""Differentiable dummy discriminator for StarGAN."""
hidden = tf.compat.v1.layers.flatten(inputs)
output_src = tf.reduce_mean(input_tensor=hidden, axis=1)
output_cls = tf.compat.v1.layers.dense(hidden, num_domains)
return output_src, output_cls
with tf.compat.v1.variable_scope('discriminator') as dis_scope:
pass
self.model = tfgan.StarGANModel(
input_data=self.input_data,
input_data_domain_label=self.input_data_domain_label,
generated_data=self.generated_data,
generated_data_domain_target=None,
reconstructed_data=None,
discriminator_input_data_source_predication=self.
discriminator_input_data_source_predication,
discriminator_generated_data_source_predication=self.
discriminator_generated_data_source_predication,
discriminator_input_data_domain_predication=None,
discriminator_generated_data_domain_predication=None,
generator_variables=None,
generator_scope=None,
generator_fn=None,
discriminator_variables=None,
discriminator_scope=dis_scope,
discriminator_fn=_discriminator_fn)
self.discriminator_fn = _discriminator_fn
self.discriminator_scope = dis_scope
def get_stargan_model():
"""Similar to get_gan_model()."""
# TODO(joelshor): Find a better way of creating a variable scope.
with tf.compat.v1.variable_scope('discriminator') as dis_scope:
pass
with tf.compat.v1.variable_scope('generator') as gen_scope:
return tfgan.StarGANModel(
input_data=tf.ones([1, 2, 2, 3]),
input_data_domain_label=tf.ones([1, 2]),
generated_data=stargan_generator_model(tf.ones([1, 2, 2, 3]),
None),
generated_data_domain_target=tf.ones([1, 2]),
reconstructed_data=tf.ones([1, 2, 2, 3]),
discriminator_input_data_source_predication=tf.ones([1]),
discriminator_generated_data_source_predication=tf.ones([1]),
discriminator_input_data_domain_predication=tf.ones([1, 2]),
discriminator_generated_data_domain_predication=tf.ones([1, 2]),
generator_variables=None,
generator_scope=gen_scope,
generator_fn=stargan_generator_model,
discriminator_variables=None,
discriminator_scope=dis_scope,
discriminator_fn=discriminator_model)