import tensorflow as tf
import tensorflow.contrib.layers as ly
import os, sys
sys.path.append(os.path.abspath(os.path.join(os.getcwd(), "../../")))
from aae import Config, Operation
'''
build layers
'''
config = Config()
config.num_types_of_label = 10
opt = Operation()
def encoder_x_z(x):
img = ly.fully_connected(x,
1000,
activation_fn=config.nonlinearity,
normalizer_fn=ly.batch_norm,
normalizer_params={'fused': True},
weights_initializer=tf.random_normal_initializer(
0, 0.01))
img = ly.fully_connected(img,
1000,
activation_fn=config.nonlinearity,
normalizer_fn=ly.batch_norm,
normalizer_params={'fused': True},
weights_initializer=tf.random_normal_initializer(
import tensorflow as tf
import tensorflow.contrib.layers as ly
import os, sys
sys.path.append(os.path.abspath(os.path.join(os.getcwd(), "../../")))
from aae import Config, Operation
'''
build layers
'''
config = Config()
config.num_types_of_label = 10
config.distribution_z = 'gaussian'
opt = Operation()
def encoder_x_z(x):
img = ly.fully_connected(x, 1000, activation_fn=config.nonlinearity, normalizer_fn=ly.batch_norm,
normalizer_params={'fused': True},
weights_initializer=tf.random_normal_initializer(0, 0.01))
img = ly.fully_connected(img, 1000, activation_fn=config.nonlinearity, normalizer_fn=ly.batch_norm,
normalizer_params={'fused': True},
weights_initializer=tf.random_normal_initializer(0, 0.01))
if config.distribution_z == 'deterministic':
img = ly.fully_connected(img, config.ndim_z, activation_fn=None,
weights_initializer=tf.random_normal_initializer(0, 0.01))
elif config.distribution_z == 'gaussian':
mu = ly.fully_connected(img, config.ndim_z, activation_fn=None,
weights_initializer=tf.random_normal_initializer(0, 0.01))
import tensorflow as tf
import tensorflow.contrib.layers as ly
import os, sys
sys.path.append(os.path.abspath(os.path.join(os.getcwd(), "../../")))
from aae import Config, Operation
'''
build layers
'''
config = Config()
opt = Operation()
def encoder_x_z(x):
img = ly.fully_connected(x,
1000,
activation_fn=config.nonlinearity,
normalizer_fn=ly.batch_norm,
normalizer_params={'fused': True},
weights_initializer=tf.random_normal_initializer(
0, 0.01))
img = ly.fully_connected(img,
1000,
activation_fn=config.nonlinearity,
normalizer_fn=ly.batch_norm,
normalizer_params={'fused': True},
weights_initializer=tf.random_normal_initializer(
0, 0.01))
import tensorflow as tf
import tensorflow.contrib.layers as ly
import os, sys
sys.path.append(os.path.abspath(os.path.join(os.getcwd(), "../../")))
from aae import Config, Operation
'''
build layers
'''
config = Config()
config.num_types_of_label = 10
config.distribution_z = 'deterministic'
config.distribution_z_adversarial = True
config.ndim_noise = 30
config.distribution_sampler = 'gaussian_mixture'
opt = Operation()
def encoder_x_z(x, z):
img = tf.concat([z, x], axis=-1)
img = ly.fully_connected(img, 1000, activation_fn=config.nonlinearity, normalizer_fn=ly.batch_norm,
normalizer_params={'fused': True},
weights_initializer=tf.random_normal_initializer(0, 0.01))
img = ly.fully_connected(img, 1000, activation_fn=config.nonlinearity, normalizer_fn=ly.batch_norm,
normalizer_params={'fused': True},
weights_initializer=tf.random_normal_initializer(0, 0.01))
if config.distribution_z == 'deterministic':
img = ly.fully_connected(img, config.ndim_z, activation_fn=None,