def check_split_latent_conditioning(self, merge_std):
with tf.Graph().as_default():
rng = np.random.RandomState(0)
x_rand = rng.randn(12, 32, 32, 32).astype(np.float32)
latent_rand = rng.randn(12, 32, 32, 16).astype(np.float32)
x_t = tf.convert_to_tensor(x_rand)
latent_t = tf.convert_to_tensor(latent_rand)
hparams = glow.glow_hparams()
hparams.level_scale = merge_std
hparams.add_hparam("latent_dist_encoder", "pointwise")
# Test initalization.
# x2 ~ N(scale * latent, 1.0) where initial scale is 1.0
exp_x2 = x_rand[:, :, :, 16:]
exp_eps = x_rand[:, :, :, 16:] - latent_rand
x_inv, _, eps, x2_t, _ = glow_ops.split(
merge_std, x_t, cond_latents=latent_t, hparams=hparams,
condition=True)
# Test reversibility.
x_inv_inv, _, _ = glow_ops.split(
merge_std, x_inv, cond_latents=latent_t, eps=eps, reverse=True,
hparams=hparams, condition=True)
with tf.Session() as sess:
sess.run(tf.global_variables_initializer())
actual_eps, actual_x2, diff_np = sess.run([eps, x2_t, x_inv_inv - x_t])
self.assertTrue(np.allclose(diff_np, 0.0, atol=1e-5))
self.assertTrue(np.allclose(actual_eps, exp_eps))
self.assertTrue(np.allclose(exp_x2, actual_x2))
def check_split_latent_conditioning(self, merge_std):
with tf.Graph().as_default():
rng = np.random.RandomState(0)
x_rand = rng.randn(12, 32, 32, 32).astype(np.float32)
latent_rand = rng.randn(12, 32, 32, 16).astype(np.float32)
x_t = tf.convert_to_tensor(x_rand)
latent_t = tf.convert_to_tensor(latent_rand)
hparams = glow.glow_hparams()
hparams.level_scale = merge_std
hparams.add_hparam("latent_dist_encoder", "pointwise")
# Test initalization.
# x2 ~ N(scale * latent, 1.0) where initial scale is 1.0
exp_x2 = x_rand[:, :, :, 16:]
exp_eps = x_rand[:, :, :, 16:] - latent_rand
x_inv, _, eps, x2_t, _ = glow_ops.split(
merge_std, x_t, cond_latents=latent_t, hparams=hparams,
condition=True)
# Test reversibility.
x_inv_inv, _, _ = glow_ops.split(
merge_std, x_inv, cond_latents=latent_t, eps=eps, reverse=True,
hparams=hparams, condition=True)
with tf.Session() as sess:
sess.run(tf.global_variables_initializer())
actual_eps, actual_x2, diff_np = sess.run([eps, x2_t, x_inv_inv - x_t])
self.assertTrue(np.allclose(diff_np, 0.0, atol=1e-5))
self.assertTrue(np.allclose(actual_eps, exp_eps))
self.assertTrue(np.allclose(exp_x2, actual_x2))
def test_split(self):
with tf.Graph().as_default():
x = tf.random_uniform(shape=(16, 5, 5, 32))
x_inv, _, eps = glow_ops.split("split", x)
x_inv_inv = glow_ops.split("split", x_inv, reverse=True, eps=eps)
with tf.Session() as session:
session.run(tf.global_variables_initializer())
x_inv_np, diff = session.run([x_inv, x - x_inv_inv])
self.assertEqual(x_inv_np.shape, (16, 5, 5, 16))
self.assertTrue(np.allclose(diff, 0.0, atol=1e-5))
def test_split(self):
with tf.Graph().as_default():
x = tf.random_uniform(shape=(16, 5, 5, 32))
x_inv, _, eps, z, _ = glow_ops.split("split", x)
x_inv_inv, _, _ = glow_ops.split("split", x_inv, reverse=True, eps=eps)
with tf.Session() as session:
session.run(tf.global_variables_initializer())
x_inv_np, diff, z_np = session.run([x_inv, x - x_inv_inv, z])
self.assertEqual(z_np.shape, (16, 5, 5, 16))
self.assertEqual(x_inv_np.shape, (16, 5, 5, 16))
self.assertTrue(np.allclose(diff, 0.0, atol=1e-5))