def evaluate(hparams, run_eval_loop=True):
"""Runs an evaluation loop.
Args:
hparams: An HParams instance containing the eval hyperparameters.
run_eval_loop: Whether to run the full eval loop. Set to False for testing.
"""
# Fetch real images.
with tf.compat.v1.name_scope('inputs'):
real_images, _ = data_provider.provide_data(
'train', hparams.num_images_generated, hparams.dataset_dir)
image_write_ops = None
if hparams.eval_real_images:
tf.compat.v1.summary.scalar(
'MNIST_Classifier_score',
util.mnist_score(real_images, hparams.classifier_filename))
else:
# In order for variables to load, use the same variable scope as in the
# train job.
with tf.compat.v1.variable_scope('Generator'):
images = networks.unconditional_generator(tf.random.normal(
[hparams.num_images_generated, hparams.noise_dims]),
is_training=False)
tf.compat.v1.summary.scalar(
'MNIST_Frechet_distance',
util.mnist_frechet_distance(real_images, images,
hparams.classifier_filename))
tf.compat.v1.summary.scalar(
'MNIST_Classifier_score',
util.mnist_score(images, hparams.classifier_filename))
if hparams.num_images_generated >= 100 and hparams.write_to_disk:
reshaped_images = tfgan.eval.image_reshaper(images[:100, ...],
num_cols=10)
uint8_images = data_provider.float_image_to_uint8(reshaped_images)
image_write_ops = tf.io.write_file(
'%s/%s' % (hparams.eval_dir, 'unconditional_gan.png'),
tf.image.encode_png(uint8_images[0]))
# For unit testing, use `run_eval_loop=False`.
if not run_eval_loop:
return
evaluation.evaluate_repeatedly(
hparams.checkpoint_dir,
hooks=[
evaluation.SummaryAtEndHook(hparams.eval_dir),
evaluation.StopAfterNEvalsHook(1)
],
eval_ops=image_write_ops,
max_number_of_evaluations=hparams.max_number_of_evaluations)
def _unconditional_generator(noise, mode): """MNIST generator with extra argument for tf.Estimator's `mode`.""" is_training = (mode == tf.estimator.ModeKeys.TRAIN) return networks.unconditional_generator(noise, is_training=is_training)