def optimization(self):
train_discriminator, train_generator, _ = optimizer(
self.beta_1,
self.loss_gen,
self.loss_dis,
self.loss_type,
self.learning_rate_input_d,
self.learning_rate_input_g,
None,
beta_2=self.beta_2,
gen_name='generator',
dis_name='discriminator_gen',
mapping_name='mapping_',
encoder_name='encoder')
with tf.control_dependencies(tf.get_collection(
tf.GraphKeys.UPDATE_OPS)):
# Quick dirty optimizer for Encoder.
trainable_variables = tf.trainable_variables()
encoder_variables = [
variable for variable in trainable_variables
if variable.name.startswith('encoder')
]
train_encoder = tf.train.AdamOptimizer(
learning_rate=self.learning_rate_input_e,
beta1=self.beta_1).minimize(self.loss_enc,
var_list=encoder_variables)
return train_discriminator, train_generator, train_encoder
def optimization(self):
train_discriminator, train_generator = optimizer(
self.beta_1,
self.loss_gen,
self.loss_dis,
self.loss_type,
self.learning_rate_input_g,
self.learning_rate_input_d,
beta_2=self.beta_2)
return train_discriminator, train_generator
def optimization(self): train_discriminator, train_generator = optimizer(self.beta_1, self.loss_gen, self.loss_dis, self.loss_type, self.learning_rate_input_g, self.learning_rate_input_d, beta_2=self.beta_2) return train_discriminator, train_generator
