def test_apply_normalization():
input_shape = (1, 4)
reshaped_inputs = tf.constant([[[2.0, 2.0], [3.0, 3.0]]])
layer = GroupNormalization(groups=2, axis=1, scale=False, center=False)
normalized_input = layer._apply_normalization(reshaped_inputs, input_shape)
np.testing.assert_equal(normalized_input,
np.array([[[0.0, 0.0], [0.0, 0.0]]]))
def test_apply_normalization(self):
input_shape = (1, 4)
expected_shape = (1, 2, 2)
reshaped_inputs = tf.constant([[[2.0, 2.0], [3.0, 3.0]]])
layer = GroupNormalization(groups=2, axis=1, scale=False, center=False)
normalized_input = layer._apply_normalization(reshaped_inputs,
input_shape)
self.assertTrue(
np.all(
np.equal(self.evaluate(normalized_input),
np.array([[[0.0, 0.0], [0.0, 0.0]]]))))
