def main(_):
if not tf.gfile.Exists(FLAGS.train_log_dir):
tf.gfile.MakeDirs(FLAGS.train_log_dir)
with tf.name_scope('inputs'):
with tf.device('/cpu:0'):
images_vae, one_hot_labels, _ = provide_data('train', FLAGS.batch_size, FLAGS.dataset_dir, num_threads=4)
images_gan = 2.0 * images_vae - 1.0
my_vae = VAE("train", z_dim=64, data_tensor=images_vae)
rec = my_vae.reconstruct(images_vae)
vae_checkpoint_path = tf.train.latest_checkpoint(FLAGS.vae_checkpoint_folder)
saver = tf.train.Saver()
gan_model = tfgan.gan_model(
generator_fn=networks.generator,
discriminator_fn=networks.discriminator,
real_data=images_gan,
generator_inputs=[tf.random_normal(
[FLAGS.batch_size, FLAGS.noise_dims]), tf.reshape(rec, [FLAGS.batch_size, 28, 28, 1])])
tfgan.eval.add_gan_model_image_summaries(gan_model, FLAGS.grid_size, True)
with tf.name_scope('loss'):
gan_loss = tfgan.gan_loss(
gan_model,
gradient_penalty_weight=1.0,
mutual_information_penalty_weight=0.0,
add_summaries=True)
# tfgan.eval.add_regularization_loss_summaries(gan_model)
# Get the GANTrain ops using custom optimizers.
with tf.name_scope('train'):
gen_lr, dis_lr = (1e-3, 1e-4)
train_ops = tfgan.gan_train_ops(
gan_model,
gan_loss,
generator_optimizer=tf.train.AdamOptimizer(gen_lr, 0.5),
discriminator_optimizer=tf.train.AdamOptimizer(dis_lr, 0.5),
summarize_gradients=False,
aggregation_method=tf.AggregationMethod.EXPERIMENTAL_ACCUMULATE_N)
status_message = tf.string_join(
['Starting train step: ',
tf.as_string(tf.train.get_or_create_global_step())],
name='status_message')
step_hooks = tfgan.get_sequential_train_hooks()(train_ops)
hooks = [tf.train.StopAtStepHook(num_steps=FLAGS.max_number_of_steps),
tf.train.LoggingTensorHook([status_message], every_n_iter=10)] + list(step_hooks)
with tf.train.MonitoredTrainingSession(hooks=hooks,
save_summaries_steps=500,
checkpoint_dir=FLAGS.train_log_dir) as sess:
saver.restore(sess, vae_checkpoint_path)
loss = None
while not sess.should_stop():
loss = sess.run(train_ops.global_step_inc_op)
def get_model_and_loss(condition, real_image):
gan_model = tfgan.gan_model(generator_fn=generator_fn,
discriminator_fn=discriminator_fn,
real_data=real_image,
generator_inputs=condition)
gan_loss = tfgan.gan_loss(gan_model,
generator_loss_fn=generator_loss_fn,
discriminator_loss_fn=discriminator_loss_fn)
return gan_model, gan_loss
def train_noestimator(features,
labels,
noise_dims=64,
batch_size=32,
num_steps=1200,
num_eval=20,
seed=0):
""" Input features (images) and labels, noise vector dimension, batch size, seed for reproducibility """
# Input training data and noise
train_input_fn, train_input_hook = \
_get_train_input_fn(features, labels, batch_size, noise_dims, seed)
noise, next_image_batch = train_input_fn()
# Define GAN model, loss, and optimizers
model = tfgan.gan_model(generator_fn, discriminator_fn, next_image_batch,
noise)
loss = tfgan.gan_loss(
model,
generator_loss_fn=tfgan.losses.wasserstein_generator_loss,
discriminator_loss_fn=tfgan.losses.wasserstein_discriminator_loss,
gradient_penalty_weight=1.0)
generator_optimizer = tf.train.AdamOptimizer(0.001, beta1=0.5)
discriminator_optimizer = tf.train.AdamOptimizer(0.0001, beta1=0.5)
gan_train_ops = tfgan.gan_train_ops(model, loss, generator_optimizer,
discriminator_optimizer)
# We'll evaluate images during training to see how the generator improves
with tf.variable_scope('Generator', reuse=True):
predict_input_fn = _get_predict_input_fn(num_eval, noise_dims)
eval_images = model.generator_fn(predict_input_fn(), is_training=False)
# Train, outputting evaluation occasionally
train_step_fn = tfgan.get_sequential_train_steps()
global_step = tf.train.get_or_create_global_step()
with tf.train.SingularMonitoredSession(hooks=[train_input_hook]) as sess:
for i in range(num_steps + 1):
cur_loss, _ = train_step_fn(sess,
gan_train_ops,
global_step,
train_step_kwargs={})
if i % 400 == 0:
generated_images = sess.run(eval_images)
print("Iteration", i, "- Loss:", cur_loss)
show(generated_images)
def model_fn(features, labels, mode, params):
if mode == tf.estimator.ModeKeys.PREDICT:
raise NotImplementedError()
else:
# Pull images from input
x = features['x']
# Generate latent samples of same batch size as images
n = tf.shape(x)[0]
rnd = tf.random_normal(shape=(n, params.latent_units),
mean=0.,
stddev=1.,
dtype=tf.float32)
# Build GAN Model
gan_model = tfgan.gan_model(generator_fn=generator_fn,
discriminator_fn=discriminator_fn,
real_data=x,
generator_inputs=rnd)
gan_loss = tfgan.gan_loss(
gan_model,
generator_loss_fn=tfgan.losses.modified_generator_loss,
discriminator_loss_fn=tfgan.losses.modified_discriminator_loss)
if mode == tf.estimator.ModeKeys.TRAIN:
generate_grid(gan_model, params)
train_ops = tfgan.gan_train_ops(
gan_model,
gan_loss,
generator_optimizer=tf.train.RMSPropOptimizer(params.gen_lr),
discriminator_optimizer=tf.train.RMSPropOptimizer(
params.dis_lr))
gan_hooks = tfgan.get_sequential_train_hooks(
GANTrainSteps(params.generator_steps,
params.discriminator_steps))(train_ops)
return tf.estimator.EstimatorSpec(
mode=mode,
loss=gan_loss.discriminator_loss,
train_op=train_ops.global_step_inc_op,
training_hooks=gan_hooks)
else:
eval_metric_ops = {}
return tf.estimator.EstimatorSpec(mode=mode,
loss=gan_loss.discriminator_loss,
eval_metric_ops=eval_metric_ops)
def aegan_model(
# Lambdas defining models.
generator_fn,
discriminator_fn,
encoder_fn,
# Real data and conditioning.
real_data,
generator_inputs,
# Optional scopes.
generator_scope='Generator',
discriminator_scope='Discriminator',
encoder_scope='Encoder',
# Options.
check_shapes=True):
gan_model = tfgan.gan_model(generator_fn,
discriminator_fn,
real_data,
generator_inputs,
generator_scope=generator_scope,
discriminator_scope=discriminator_scope,
check_shapes=check_shapes)
with tf.variable_scope(encoder_scope) as enc_scope:
encoder_gen_outputs = encoder_fn(gan_model.generated_data)
with tf.variable_scope(enc_scope, reuse=True):
real_data = tf.convert_to_tensor(real_data)
encoder_real_outputs = encoder_fn(real_data)
encoder_variables = tf.trainable_variables(scope=encoder_scope)
return AEGANModel(
generator_inputs, gan_model.generated_data,
gan_model.generator_variables, gan_model.generator_scope, generator_fn,
real_data, gan_model.discriminator_real_outputs,
gan_model.discriminator_gen_outputs, gan_model.discriminator_variables,
gan_model.discriminator_scope, discriminator_fn, encoder_real_outputs,
encoder_gen_outputs, encoder_variables, enc_scope, encoder_fn)
def build_gan_harness(image_input: tf.Tensor,
noise: tf.Tensor,
generator: tf.keras.Model,
discriminator: tf.keras.Model,
generator_learning_rate=0.01,
discriminator_learning_rate=0.01,
noise_format: str = 'SPHERE',
adversarial_training: str = 'WASSERSTEIN',
feature_matching: bool = False,
no_trainer: bool = False,
summarize_activations: bool = False) -> tuple:
image_size = image_input.shape.as_list()[1]
nchannels = image_input.shape.as_list()[3]
print("Plain Generative Adversarial Network: {}x{}x{} images".format(
image_size, image_size, nchannels))
def _generator_fn(z):
return generator([z], training=True)
def _discriminator_fn(x, z):
return discriminator([x, z], training=True)
gan_model = tfgan.gan_model(
_generator_fn,
_discriminator_fn,
image_input,
noise,
generator_scope='Generator',
discriminator_scope='Discriminator',
check_shapes=True) # set to False for 2-level architectures
sampled_x = gan_model.generated_data
image_grid_summary(sampled_x, grid_size=3, name='generated_data')
if summarize_activations:
tf.contrib.layers.summarize_activations()
tf.contrib.layers.summarize_collection(tf.GraphKeys.TRAINABLE_VARIABLES)
loss = gan_loss_by_name(gan_model,
adversarial_training,
feature_matching=feature_matching,
add_summaries=True)
if adversarial_training != 'WASSERSTEIN' and adversarial_training != 'RELATIVISTIC_AVG':
disc_accuracy_gen = basic_accuracy(
tf.zeros_like(gan_model.discriminator_gen_outputs),
gan_model.discriminator_gen_outputs)
disc_accuracy_real = basic_accuracy(
tf.ones_like(gan_model.discriminator_real_outputs),
gan_model.discriminator_real_outputs)
disc_accuracy = (disc_accuracy_gen + disc_accuracy_real) * 0.5
with tf.name_scope('Discriminator'):
tf.summary.scalar('accuracy', disc_accuracy)
if no_trainer:
train_ops = None
else:
train_ops = tfgan.gan_train_ops(
gan_model,
loss,
generator_optimizer=tf.train.AdamOptimizer(generator_learning_rate,
beta1=0.,
beta2=0.99),
discriminator_optimizer=tf.train.AdamOptimizer(
discriminator_learning_rate, beta1=0., beta2=0.99),
summarize_gradients=True)
return (gan_model, loss, train_ops)
def run_discgan():
""" Constructs and trains the discriminative GAN consisting of
Jerry and Diego.
"""
# code follows the examples from
# https://github.com/tensorflow/models/blob/master/research/gan/tutorial.ipynb
# build the GAN model
discgan = tfgan.gan_model(
generator_fn=generator,
discriminator_fn=adversary_conv(OUTPUT_SIZE),
real_data=tf.random_uniform(shape=[BATCH_SIZE, OUTPUT_SIZE]),
generator_inputs=get_input_tensor(BATCH_SIZE, MAX_VAL))
# Build the GAN loss
discgan_loss = tfgan.gan_loss(
discgan,
generator_loss_fn=tfgan.losses.least_squares_generator_loss,
discriminator_loss_fn=tfgan.losses.least_squares_discriminator_loss)
# Create the train ops, which calculate gradients and apply updates to weights.
train_ops = tfgan.gan_train_ops(discgan,
discgan_loss,
generator_optimizer=GEN_OPT,
discriminator_optimizer=OPP_OPT)
# start TensorFlow session
with tf.train.SingularMonitoredSession() as sess:
pretrain_steps_fn = tfgan.get_sequential_train_steps(
tfgan.GANTrainSteps(0, PRE_STEPS))
train_steps_fn = tfgan.get_sequential_train_steps(
tfgan.GANTrainSteps(1, ADV_MULT))
global_step = tf.train.get_or_create_global_step()
# pretrain discriminator
print('\n\nPretraining ... ', end="", flush=True)
try:
pretrain_steps_fn(sess,
train_ops,
global_step,
train_step_kwargs={})
except KeyboardInterrupt:
pass
print('[DONE]\n\n')
# train both models
losses_jerry = []
losses_diego = []
try:
evaluate(sess, discgan.generated_data, discgan.generator_inputs, 0,
'jerry')
for step in range(STEPS):
train_steps_fn(sess,
train_ops,
global_step,
train_step_kwargs={})
# if performed right number of steps, log
if step % LOG_EVERY_N == 0:
sess.run([])
gen_l = discgan_loss.generator_loss.eval(session=sess)
disc_l = discgan_loss.discriminator_loss.eval(session=sess)
debug.print_step(step, gen_l, disc_l)
losses_jerry.append(gen_l)
losses_diego.append(disc_l)
except KeyboardInterrupt:
print('[INTERRUPTED BY USER] -- evaluating')
# produce output
files.write_to_file(losses_jerry, PLOT_DIR + '/jerry_loss.txt')
files.write_to_file(losses_diego, PLOT_DIR + '/diego_loss.txt')
evaluate(sess, discgan.generated_data, discgan.generator_inputs, 1,
'jerry')
weights_regularizer=layers.l2_regularizer(weight_decay),
biases_regularizer=layers.l2_regularizer(weight_decay)):
net = layers.fully_connected(fragment, 64)
net = layers.dropout(net, keep_prob=0.75)
net = layers.fully_connected(net, 32)
net = layers.fully_connected(net, 16, normalizer_fn=layers.batch_norm,activation_fn=tf.tanh)
return layers.linear(net, 1, normalizer_fn=None,activation_fn=tf.tanh)
real_data_normed = tf.divide(tf.convert_to_tensor(real, dtype=tf.float32), tf.constant(MAX_VAL, dtype=tf.float32))
chunk_queue = tf.train.slice_input_producer([real_data_normed])
# Build the generator and discriminator.
gan_model = tfgan.gan_model(
generator_fn=generator_fn, # you define
discriminator_fn=discriminator_fn, # you define
real_data=chunk_queue,
generator_inputs=noise)
gan_loss = tfgan.gan_loss(
gan_model,
generator_loss_fn=tfgan.losses.wasserstein_generator_loss,
discriminator_loss_fn=tfgan.losses.wasserstein_discriminator_loss,
gradient_penalty_weight=1.0)
l1_loss = tf.norm(gan_model.real_data - gan_model.generated_data, ord=1)
gan_loss = tfgan.losses.combine_adversarial_loss(gan_loss, gan_model, l1_loss, weight_factor=FLAGS.weight_factor)
train_ops = tfgan.gan_train_ops(gan_model,gan_loss,generator_optimizer=tf.train.AdamOptimizer(learning_rate=0.001, beta1=0.85, beta2=0.999, epsilon=1e-5),discriminator_optimizer=tf.train.AdamOptimizer(learning_rate=0.000001, beta1=0.85, beta2=0.999, epsilon=1e-5))
#train_ops.global_step_inc_op = tf.train.get_global_step().assign_add(1)
biases_regularizer=layers.l2_regularizer(weight_decay)):
net = layers.conv2d(img, 64, [4, 4], stride=2)
net = tfgan.features.condition_tensor_from_onehot(net, one_hot_labels)
net = layers.conv2d(net, 128, [4, 4], stride=2)
net = layers.flatten(net)
net = layers.fully_connected(net,
1024,
normalizer_fn=layers.batch_norm)
return layers.linear(net, 1)
noise_dims = 64
conditional_gan_model = tfgan.gan_model(
generator_fn=conditional_generator_fn,
discriminator_fn=conditional_discriminator_fn,
real_data=real_images,
generator_inputs=(tf.random_normal([batch_size,
noise_dims]), one_hot_labels))
# Sanity check that currently generated images are garbage.
cond_generated_data_to_visualize = tfgan.eval.image_reshaper(
conditional_gan_model.generated_data[:20, ...], num_cols=10)
visualize_digits(cond_generated_data_to_visualize)
gan_loss = tfgan.gan_loss(conditional_gan_model, gradient_penalty_weight=1.0)
# Sanity check that we can evaluate our losses.
evaluate_tfgan_loss(gan_loss)
generator_optimizer = tf.train.AdamOptimizer(0.0009, beta1=0.5)
discriminator_optimizer = tf.train.AdamOptimizer(0.00009, beta1=0.5)
def model_fn(features, labels, mode, params):
is_chief = not tf.get_variable_scope().reuse
batch_size = tf.shape(labels)[0]
noise = tf.random_normal([batch_size, FLAGS.emb_dim])
noise = tf.nn.l2_normalize(noise, axis=1)
gan_model = tfgan.gan_model(generator_fn=generator,
discriminator_fn=discriminator,
real_data=features[:, 1:],
generator_inputs=(noise, labels - 1),
check_shapes=False)
if is_chief:
for variable in tf.trainable_variables():
tf.summary.histogram(variable.op.name, variable)
tf.summary.histogram('logits/gen_logits',
gan_model.discriminator_gen_outputs[0])
tf.summary.histogram('logits/real_logits',
gan_model.discriminator_real_outputs[0])
def gen_loss_fn(gan_model, add_summaries):
return 0
def dis_loss_fn(gan_model, add_summaries):
discriminator_real_outputs = gan_model.discriminator_real_outputs
discriminator_gen_outputs = gan_model.discriminator_gen_outputs
real_logits = tf.boolean_mask(discriminator_real_outputs[0],
discriminator_real_outputs[1])
gen_logits = tf.boolean_mask(discriminator_gen_outputs[0],
discriminator_gen_outputs[1])
return modified_discriminator_loss(real_logits,
gen_logits,
add_summaries=add_summaries)
with tf.name_scope('losses'):
gan_loss = tfgan.gan_loss(
gan_model,
generator_loss_fn=gen_loss_fn,
discriminator_loss_fn=dis_loss_fn,
gradient_penalty_weight=10 if FLAGS.wass else 0,
add_summaries=is_chief)
if is_chief:
tfgan.eval.add_regularization_loss_summaries(gan_model)
gan_loss = rl_loss(gan_model, gan_loss, add_summaries=is_chief)
loss = gan_loss.generator_loss + gan_loss.discriminator_loss
with tf.name_scope('train'):
gen_opt = tf.train.AdamOptimizer(params.gen_lr, 0.5)
dis_opt = tf.train.AdamOptimizer(params.dis_lr, 0.5)
if params.multi_gpu:
gen_opt = tf.contrib.estimator.TowerOptimizer(gen_opt)
dis_opt = tf.contrib.estimator.TowerOptimizer(dis_opt)
train_ops = tfgan.gan_train_ops(
gan_model,
gan_loss,
generator_optimizer=gen_opt,
discriminator_optimizer=dis_opt,
transform_grads_fn=transform_grads_fn,
summarize_gradients=is_chief,
check_for_unused_update_ops=True,
aggregation_method=tf.AggregationMethod.EXPERIMENTAL_ACCUMULATE_N)
train_op = train_ops.global_step_inc_op
train_hooks = get_sequential_train_hooks()(train_ops)
if is_chief:
with open('data/word_counts.txt', 'r') as f:
dic = list(f)
dic = [i.split()[0] for i in dic]
dic.append('<unk>')
dic = tf.convert_to_tensor(dic)
sentence = crop_sentence(gan_model.generated_data[0][0], FLAGS.end_id)
sentence = tf.gather(dic, sentence)
real = crop_sentence(gan_model.real_data[0], FLAGS.end_id)
real = tf.gather(dic, real)
train_hooks.append(
tf.train.LoggingTensorHook({
'fake': sentence,
'real': real
},
every_n_iter=100))
tf.summary.text('fake', sentence)
gen_var = tf.trainable_variables('Generator')
dis_var = []
dis_var.extend(tf.trainable_variables('Discriminator/rnn'))
dis_var.extend(tf.trainable_variables('Discriminator/embedding'))
saver = tf.train.Saver(gen_var + dis_var)
def init_fn(scaffold, session):
saver.restore(session, FLAGS.sae_ckpt)
pass
scaffold = tf.train.Scaffold(init_fn=init_fn)
return tf.estimator.EstimatorSpec(mode=mode,
loss=loss,
train_op=train_op,
scaffold=scaffold,
training_hooks=train_hooks)
weights_regularizer=layers.l2_regularizer(weight_decay),
biases_regularizer=layers.l2_regularizer(weight_decay)):
net = layers.conv2d(img, 64, [4, 4], stride=2)
net = layers.conv2d(net, 128, [4, 4], stride=2)
net = layers.flatten(net)
with framework.arg_scope([layers.batch_norm], is_training=is_training):
net = layers.fully_connected(net,
1024,
normalizer_fn=layers.batch_norm)
return layers.linear(net, 1)
noise_dims = 64
gan_model = tfgan.gan_model(generator_fn,
discriminator_fn,
real_data=real_images,
generator_inputs=tf.random_normal(
[batch_size, noise_dims]))
# Sanity check that generated images before training are garbage.
check_generated_digits = tfgan.eval.image_reshaper(
gan_model.generated_data[:20, ...], num_cols=10)
visualize_digits(check_generated_digits)
# We can use the minimax loss from the original paper.
vanilla_gan_loss = tfgan.gan_loss(
gan_model,
generator_loss_fn=tfgan.losses.minimax_generator_loss,
discriminator_loss_fn=tfgan.losses.minimax_discriminator_loss)
# We can use the Wasserstein loss (https://arxiv.org/abs/1701.07875) with the
def model_fn(features, labels, mode, params):
"""The full unsupervised captioning model."""
is_chief = not tf.get_variable_scope().reuse
with slim.arg_scope(inception_v4.inception_v4_arg_scope()):
net, _ = inception_v4.inception_v4(features['im'],
None,
is_training=False)
net = tf.squeeze(net, [1, 2])
inc_saver = tf.train.Saver(tf.global_variables('InceptionV4'))
gan_model = tfgan.gan_model(generator_fn=generator,
discriminator_fn=discriminator,
real_data=labels['sentence'][:, 1:],
generator_inputs=(net, labels['len'] - 1),
check_shapes=False)
if is_chief:
for variable in tf.trainable_variables():
tf.summary.histogram(variable.op.name, variable)
tf.summary.histogram('logits/gen_logits',
gan_model.discriminator_gen_outputs[0])
tf.summary.histogram('logits/real_logits',
gan_model.discriminator_real_outputs[0])
def gen_loss_fn(gan_model, add_summaries):
return 0
def dis_loss_fn(gan_model, add_summaries):
discriminator_real_outputs = gan_model.discriminator_real_outputs
discriminator_gen_outputs = gan_model.discriminator_gen_outputs
real_logits = tf.boolean_mask(discriminator_real_outputs[0],
discriminator_real_outputs[1])
gen_logits = tf.boolean_mask(discriminator_gen_outputs[0],
discriminator_gen_outputs[1])
return modified_discriminator_loss(real_logits,
gen_logits,
add_summaries=add_summaries)
with tf.name_scope('losses'):
pool_fn = functools.partial(tfgan.features.tensor_pool,
pool_size=FLAGS.pool_size)
gan_loss = tfgan.gan_loss(
gan_model,
generator_loss_fn=gen_loss_fn,
discriminator_loss_fn=dis_loss_fn,
gradient_penalty_weight=10 if FLAGS.wass else 0,
tensor_pool_fn=pool_fn if FLAGS.use_pool else None,
add_summaries=is_chief)
if is_chief:
tfgan.eval.add_regularization_loss_summaries(gan_model)
gan_loss = rl_loss(gan_model,
gan_loss,
features['classes'],
features['scores'],
features['num'],
add_summaries=is_chief)
sen_ae_loss = sentence_ae(gan_model, features, labels, is_chief)
loss = gan_loss.generator_loss + gan_loss.discriminator_loss + sen_ae_loss
gan_loss = gan_loss._replace(generator_loss=gan_loss.generator_loss +
sen_ae_loss)
with tf.name_scope('train'):
gen_opt = tf.train.AdamOptimizer(params.gen_lr, 0.5)
dis_opt = tf.train.AdamOptimizer(params.dis_lr, 0.5)
if params.multi_gpu:
gen_opt = tf.contrib.estimator.TowerOptimizer(gen_opt)
dis_opt = tf.contrib.estimator.TowerOptimizer(dis_opt)
train_ops = tfgan.gan_train_ops(
gan_model,
gan_loss,
generator_optimizer=gen_opt,
discriminator_optimizer=dis_opt,
transform_grads_fn=transform_grads_fn,
summarize_gradients=is_chief,
check_for_unused_update_ops=not FLAGS.use_pool,
aggregation_method=tf.AggregationMethod.EXPERIMENTAL_ACCUMULATE_N)
train_op = train_ops.global_step_inc_op
train_hooks = get_sequential_train_hooks()(train_ops)
# Summary the generated caption on the fly.
if is_chief:
with open('data/word_counts.txt', 'r') as f:
dic = list(f)
dic = [i.split()[0] for i in dic]
dic.append('<unk>')
dic = tf.convert_to_tensor(dic)
sentence = crop_sentence(gan_model.generated_data[0][0], FLAGS.end_id)
sentence = tf.gather(dic, sentence)
real = crop_sentence(gan_model.real_data[0], FLAGS.end_id)
real = tf.gather(dic, real)
train_hooks.append(
tf.train.LoggingTensorHook({
'fake': sentence,
'real': real
},
every_n_iter=100))
tf.summary.text('fake', sentence)
tf.summary.image('im', features['im'][None, 0])
gen_saver = tf.train.Saver(tf.trainable_variables('Generator'))
dis_var = []
dis_var.extend(tf.trainable_variables('Discriminator/rnn'))
dis_var.extend(tf.trainable_variables('Discriminator/embedding'))
dis_var.extend(tf.trainable_variables('Discriminator/fc'))
dis_saver = tf.train.Saver(dis_var)
def init_fn(scaffold, session):
inc_saver.restore(session, FLAGS.inc_ckpt)
if FLAGS.imcap_ckpt:
gen_saver.restore(session, FLAGS.imcap_ckpt)
if FLAGS.sae_ckpt:
dis_saver.restore(session, FLAGS.sae_ckpt)
scaffold = tf.train.Scaffold(init_fn=init_fn)
return tf.estimator.EstimatorSpec(mode=mode,
loss=loss,
train_op=train_op,
scaffold=scaffold,
training_hooks=train_hooks)
def main(_):
if not tf.gfile.Exists(FLAGS.train_log_dir):
tf.gfile.MakeDirs(FLAGS.train_log_dir)
with tf.name_scope('inputs'):
with tf.device('/cpu:0'):
images, one_hot_labels, _ = provide_data('train',
FLAGS.batch_size,
FLAGS.dataset_dir,
num_threads=4)
images = 2.0 * images - 1.0
gan_model = tfgan.gan_model(generator_fn=gan_networks.generator,
discriminator_fn=gan_networks.discriminator,
real_data=images,
generator_inputs=tf.random_normal(
[FLAGS.batch_size, FLAGS.noise_dims]))
tfgan.eval.add_gan_model_image_summaries(gan_model, FLAGS.grid_size, False)
with tf.variable_scope('Generator', reuse=True):
eval_images = gan_model.generator_fn(tf.random_normal(
[FLAGS.num_images_eval, FLAGS.noise_dims]),
is_training=False)
# Calculate Inception score.
tf.summary.scalar(
"Inception score",
util.mnist_score(eval_images, MNIST_CLASSIFIER_FROZEN_GRAPH))
# Calculate Frechet Inception distance.
with tf.device('/cpu:0'):
real_images, labels, _ = provide_data('train', FLAGS.num_images_eval,
FLAGS.dataset_dir)
tf.summary.scalar(
"Frechet distance",
util.mnist_frechet_distance(real_images, eval_images,
MNIST_CLASSIFIER_FROZEN_GRAPH))
with tf.name_scope('loss'):
gan_loss = tfgan.gan_loss(gan_model,
gradient_penalty_weight=1.0,
mutual_information_penalty_weight=0.0,
add_summaries=True)
# tfgan.eval.add_regularization_loss_summaries(gan_model)
with tf.name_scope('train'):
gen_lr, dis_lr = (1e-3, 1e-4)
train_ops = tfgan.gan_train_ops(
gan_model,
gan_loss,
generator_optimizer=tf.train.AdamOptimizer(gen_lr, 0.5),
discriminator_optimizer=tf.train.AdamOptimizer(dis_lr, 0.5),
summarize_gradients=False,
aggregation_method=tf.AggregationMethod.EXPERIMENTAL_ACCUMULATE_N)
status_message = tf.string_join([
'Starting train step: ',
tf.as_string(tf.train.get_or_create_global_step())
],
name='status_message')
step_hooks = tfgan.get_sequential_train_hooks()(train_ops)
hooks = [
tf.train.StopAtStepHook(num_steps=FLAGS.max_number_of_steps),
tf.train.LoggingTensorHook([status_message], every_n_iter=10)
] + list(step_hooks)
with tf.train.MonitoredTrainingSession(
hooks=hooks,
save_summaries_steps=500,
checkpoint_dir=FLAGS.train_log_dir) as sess:
loss = None
while not sess.should_stop():
loss = sess.run(train_ops.global_step_inc_op)