def testShape(self):
shape = [10, 5]
gradients = tf.random_normal(shape)
net = networks.Adam()
state = net.initial_state_for_inputs(gradients)
update, _ = net(gradients, state)
self.assertEqual(update.get_shape().as_list(), shape)
def testNonTrainable(self):
"""Tests the network doesn't contain trainable variables."""
shape = [10, 5]
gradients = tf.random_normal(shape)
net = networks.Adam()
state = net.initial_state_for_inputs(gradients)
net(gradients, state)
variables = nn.get_variables_in_module(net)
self.assertEqual(len(variables), 0)
def testZeroLearningRate(self):
"""Tests network produces zero updates with learning rate equal to zero."""
shape = [10]
gradients = tf.random_normal(shape)
net = networks.Adam(learning_rate=0)
state = net.initial_state_for_inputs(gradients)
update, _ = net(gradients, state)
with self.test_session() as sess:
update_np = sess.run(update)
self.assertAllEqual(update_np, np.zeros(shape))