def main(_):
hparams = eval_lib.HParams(FLAGS.checkpoint_dir, FLAGS.eval_dir,
FLAGS.dataset_dir, FLAGS.num_images_generated,
FLAGS.eval_real_images, FLAGS.noise_dims,
FLAGS.max_number_of_evaluations,
FLAGS.write_to_disk)
eval_lib.evaluate(hparams, run_eval_loop=True)
def test_build_graph(self, eval_real_images, mock_util, mock_provide_data):
hparams = eval_lib.HParams(
checkpoint_dir='/tmp/mnist/',
eval_dir='/tmp/mnist/',
dataset_dir=None,
num_images_generated=1000,
eval_real_images=eval_real_images,
noise_dims=64,
classifier_filename=None,
max_number_of_evaluations=None,
write_to_disk=True)
# Mock input pipeline.
bs = hparams.num_images_generated
mock_imgs = np.zeros([bs, 28, 28, 1], dtype=np.float32)
mock_lbls = np.concatenate((np.ones([bs, 1], dtype=np.int32),
np.zeros([bs, 9], dtype=np.int32)),
axis=1)
mock_provide_data.return_value = (mock_imgs, mock_lbls)
# Mock expensive eval metrics.
mock_util.mnist_frechet_distance.return_value = 1.0
mock_util.mnist_score.return_value = 0.0
eval_lib.evaluate(hparams, run_eval_loop=False)