with tf.control_dependencies(update_ops):
d_train = optimizer.minimize(d_loss, var_list=d_vars)
############ summary writing ########################################
with tf.name_scope('summaries'):
tf.summary.scalar('wasserstein_scaled', wasserstein_scaled)
tf.summary.scalar('wasserstein', wasserstein)
tf.summary.scalar('g_loss', g_loss)
tf.summary.scalar('d_loss', d_loss)
tf.summary.scalar('d_regularizer_niqe', niqe_score_mean_grad)
tf.summary.scalar('d_regularizer_gp', d_regularizer1)
tf.summary.scalar('learning_rate', learning_rate)
tf.summary.scalar('added_regularizer', added_regularizer)
tf.summary.scalar('learning_rate', learning_rate)
tf.summary.scalar('global_step', global_step)
atf.image_grid_summary('x_generated', x_generated)
merged_summary = tf.summary.merge_all()
############### intialize the variables ################
sess.run([tf.global_variables_initializer(), tf.local_variables_initializer()])
############ The image files and coordinate the loading of image files #########
coord = tf.train.Coordinator()
threads = tf.train.start_queue_runners(coord=coord)
########### Add op to save and restore #########################################
saver = tf.train.Saver(max_to_keep=10)
######## i = 1000 uncomment and enter the model number for restoring the model #####
if not reset:
nn = name + "/model.ckpt-" + str(i)
saver.restore(sess, nn)
tf.summary.scalar('d_loss', d_loss)
atf.scalars_summary('d_true', d_true)
atf.scalars_summary('d_generated', d_generated)
tf.summary.scalar('d_regularizer', d_regularizer)
tf.summary.scalar('d_regularizer_mean', d_regularizer_mean)
tf.summary.scalar('learning_rate', learning_rate)
tf.summary.scalar('global_step', global_step)
atf.scalars_summary('x_generated', x_generated)
atf.scalars_summary('x_true', x_true)
atf.scalars_summary('gamma', gamma)
atf.scalars_summary('lamb', lamb)
atf.image_grid_summary('x_true', x_true)
atf.image_grid_summary('x_generated', x_generated)
atf.image_grid_summary('gradients', gradients)
atf.image_grid_summary('dual_sobolev_gradients', dual_sobolev_gradients)
atf.scalars_summary('ddx', ddx)
atf.scalars_summary('gradients', gradients)
atf.scalars_summary('dual_sobolev_gradients', dual_sobolev_gradients)
merged_summary = tf.summary.merge_all()
# Advanced metrics
with tf.name_scope('inception'):
# Specific function to compute inception score for very large
# number of samples
def generate_and_classify(z):