Python tensorflow.contrib.data.shuffle_and_repeat() Examples
The following are 3
code examples of tensorflow.contrib.data.shuffle_and_repeat().
You can vote up the ones you like or vote down the ones you don't like,
and go to the original project or source file by following the links above each example.
You may also want to check out all available functions/classes of the module
tensorflow.contrib.data
, or try the search function
.
.
Example #1
| Source File: SphereGAN.py From SphereGAN-Tensorflow with MIT License | 4 votes |
|
def build_model(self):
""" Graph Input """
# images
Image_Data_Class = ImageData(self.img_size, self.c_dim, self.custom_dataset)
inputs = tf.data.Dataset.from_tensor_slices(self.data)
gpu_device = '/gpu:0'
inputs = inputs.\
apply(shuffle_and_repeat(self.dataset_num)).\
apply(map_and_batch(Image_Data_Class.image_processing, self.batch_size, num_parallel_batches=16, drop_remainder=True)).\
apply(prefetch_to_device(gpu_device, self.batch_size))
inputs_iterator = inputs.make_one_shot_iterator()
self.inputs = inputs_iterator.get_next()
# noises
self.z = tf.random_normal(shape=[self.batch_size, 1, 1, self.z_dim], name='random_z')
""" Loss Function """
# output of D for real images
real_logits = self.discriminator(self.inputs)
# output of D for fake images
fake_images = self.generator(self.z)
fake_logits = self.discriminator(fake_images, reuse=True)
if self.gan_type.__contains__('wgan') or self.gan_type == 'dragan':
GP = self.gradient_penalty(real=self.inputs, fake=fake_images)
else:
GP = 0
# get loss for discriminator
self.d_loss = discriminator_loss(self.gan_type, real=real_logits, fake=fake_logits, moment=self.moment) + GP
# get loss for generator
self.g_loss = generator_loss(self.gan_type, fake=fake_logits, moment=self.moment)
""" Training """
# divide trainable variables into a group for D and a group for G
t_vars = tf.trainable_variables()
d_vars = [var for var in t_vars if 'discriminator' in var.name]
g_vars = [var for var in t_vars if 'generator' in var.name]
# optimizers
with tf.control_dependencies(tf.get_collection(tf.GraphKeys.UPDATE_OPS)) :
self.d_optim = tf.train.AdamOptimizer(self.d_learning_rate, beta1=self.beta1, beta2=self.beta2).minimize(self.d_loss, var_list=d_vars)
self.g_optim = tf.train.AdamOptimizer(self.g_learning_rate, beta1=self.beta1, beta2=self.beta2).minimize(self.g_loss, var_list=g_vars)
"""" Testing """
# for test
self.fake_images = self.generator(self.z, is_training=False, reuse=True)
""" Summary """
self.d_sum = tf.summary.scalar("d_loss", self.d_loss)
self.g_sum = tf.summary.scalar("g_loss", self.g_loss)
##################################################################################
# Train
##################################################################################
Example #2
| Source File: SAGAN.py From Self-Attention-GAN-Tensorflow with MIT License | 4 votes |
|
def build_model(self):
""" Graph Input """
# images
if self.custom_dataset :
Image_Data_Class = ImageData(self.img_size, self.c_dim)
inputs = tf.data.Dataset.from_tensor_slices(self.data)
gpu_device = '/gpu:0'
inputs = inputs.apply(shuffle_and_repeat(self.dataset_num)).apply(map_and_batch(Image_Data_Class.image_processing, self.batch_size, num_parallel_batches=16, drop_remainder=True)).apply(prefetch_to_device(gpu_device, self.batch_size))
inputs_iterator = inputs.make_one_shot_iterator()
self.inputs = inputs_iterator.get_next()
else :
self.inputs = tf.placeholder(tf.float32, [self.batch_size, self.img_size, self.img_size, self.c_dim], name='real_images')
# noises
self.z = tf.placeholder(tf.float32, [self.batch_size, 1, 1, self.z_dim], name='z')
""" Loss Function """
# output of D for real images
real_logits = self.discriminator(self.inputs)
# output of D for fake images
fake_images = self.generator(self.z)
fake_logits = self.discriminator(fake_images, reuse=True)
if self.gan_type.__contains__('wgan') or self.gan_type == 'dragan' :
GP = self.gradient_penalty(real=self.inputs, fake=fake_images)
else :
GP = 0
# get loss for discriminator
self.d_loss = discriminator_loss(self.gan_type, real=real_logits, fake=fake_logits) + GP
# get loss for generator
self.g_loss = generator_loss(self.gan_type, fake=fake_logits)
""" Training """
# divide trainable variables into a group for D and a group for G
t_vars = tf.trainable_variables()
d_vars = [var for var in t_vars if 'discriminator' in var.name]
g_vars = [var for var in t_vars if 'generator' in var.name]
# optimizers
self.d_optim = tf.train.AdamOptimizer(self.d_learning_rate, beta1=self.beta1, beta2=self.beta2).minimize(self.d_loss, var_list=d_vars)
self.g_optim = tf.train.AdamOptimizer(self.g_learning_rate, beta1=self.beta1, beta2=self.beta2).minimize(self.g_loss, var_list=g_vars)
"""" Testing """
# for test
self.fake_images = self.generator(self.z, is_training=False, reuse=True)
""" Summary """
self.d_sum = tf.summary.scalar("d_loss", self.d_loss)
self.g_sum = tf.summary.scalar("g_loss", self.g_loss)
##################################################################################
# Train
##################################################################################
Example #3
| Source File: RaGAN.py From RelativisticGAN-Tensorflow with MIT License | 4 votes |
|
def build_model(self):
""" Graph Input """
# images
if self.custom_dataset :
Image_Data_Class = ImageData(self.img_size, self.c_dim)
inputs = tf.data.Dataset.from_tensor_slices(self.data)
gpu_device = '/gpu:0'
inputs = inputs.apply(shuffle_and_repeat(self.dataset_num)).apply(map_and_batch(Image_Data_Class.image_processing, self.batch_size, num_parallel_batches=16, drop_remainder=True)).apply(prefetch_to_device(gpu_device, self.batch_size))
inputs_iterator = inputs.make_one_shot_iterator()
self.inputs = inputs_iterator.get_next()
else :
self.inputs = tf.placeholder(tf.float32, [self.batch_size, self.img_size, self.img_size, self.c_dim], name='real_images')
# noises
self.z = tf.placeholder(tf.float32, [self.batch_size, 1, 1, self.z_dim], name='z')
""" Loss Function """
# output of D for real images
real_logits = self.discriminator(self.inputs)
# output of D for fake images
fake_images = self.generator(self.z)
fake_logits = self.discriminator(fake_images, reuse=True)
if self.gan_type.__contains__('gp') or self.gan_type.__contains__('lp') or self.gan_type.__contains__('dragan') :
GP = self.gradient_penalty(real=self.inputs, fake=fake_images)
else :
GP = 0
# get loss for discriminator
self.d_loss = discriminator_loss(self.Ra, self.gan_type, real=real_logits, fake=fake_logits) + GP
# get loss for generator
self.g_loss = generator_loss(self.Ra, self.gan_type, real=real_logits, fake=fake_logits)
""" Training """
# divide trainable variables into a group for D and a group for G
t_vars = tf.trainable_variables()
d_vars = [var for var in t_vars if 'discriminator' in var.name]
g_vars = [var for var in t_vars if 'generator' in var.name]
# optimizers
self.d_optim = tf.train.AdamOptimizer(self.d_learning_rate, beta1=self.beta1, beta2=self.beta2).minimize(self.d_loss, var_list=d_vars)
self.g_optim = tf.train.AdamOptimizer(self.g_learning_rate, beta1=self.beta1, beta2=self.beta2).minimize(self.g_loss, var_list=g_vars)
"""" Testing """
# for test
self.fake_images = self.generator(self.z, is_training=False, reuse=True)
""" Summary """
self.d_sum = tf.summary.scalar("d_loss", self.d_loss)
self.g_sum = tf.summary.scalar("g_loss", self.g_loss)
##################################################################################
# Train
##################################################################################
