def test_variables_receive_gradients(self):
x = tf.random.uniform((1, 5, 2), dtype=tf.float32)
layer = signal_conv.SignalConv1D(2, 3, use_bias=True)
with tf.GradientTape() as g:
y = layer(x)
grads = g.gradient(y, layer.trainable_weights)
self.assertLen(grads, 2)
self.assertNotIn(None, grads)
grad_shapes = [tuple(g.shape) for g in grads]
weight_shapes = [tuple(w.shape) for w in layer.trainable_weights]
self.assertSameElements(grad_shapes, weight_shapes)
def test_can_be_saved_within_functional_model(self, build):
inputs = tf.keras.Input(shape=(None, 2))
outputs = signal_conv.SignalConv1D(1,
3,
use_bias=True,
activation=tf.nn.relu)(inputs)
model = tf.keras.Model(inputs=inputs, outputs=outputs)
layer = model.get_layer("signal_conv1d")
with self.subTest(name="layer_created_as_expected"):
self.assertIsInstance(layer, signal_conv.SignalConv1D)
self.assertIsInstance(layer.kernel_parameter,
parameters.RDFTParameter)
self.assertIsInstance(layer.bias_parameter, tf.Variable)
if build:
x = tf.random.uniform((1, 5, 2), dtype=tf.float32)
y = model(x)
weight_names = [w.name for w in model.weights]
tempdir = self.create_tempdir()
model_path = os.path.join(tempdir, "model")
# This should force the model to be reconstructed via configs.
model.save(model_path, save_traces=False)
model = tf.keras.models.load_model(model_path)
layer = model.get_layer("signal_conv1d")
with self.subTest(name="layer_recreated_as_expected"):
self.assertIsInstance(layer, signal_conv.SignalConv1D)
self.assertIsInstance(layer.kernel_parameter,
parameters.RDFTParameter)
self.assertIsInstance(layer.bias_parameter, tf.Variable)
if build:
with self.subTest(name="model_outputs_identical"):
self.assertAllEqual(model(x), y)
with self.subTest(name="model_weights_identical"):
self.assertSameElements(weight_names,
[w.name for w in model.weights])
def test_attributes_cannot_be_set_after_build(self):
layer = signal_conv.SignalConv1D(2, 1)
layer.build((None, None, 2))
with self.assertRaises(RuntimeError):
layer.filters = 3
with self.assertRaises(RuntimeError):
layer.kernel_support = 3
with self.assertRaises(RuntimeError):
layer.corr = True
with self.assertRaises(RuntimeError):
layer.strides_down = 2
with self.assertRaises(RuntimeError):
layer.strides_up = 2
with self.assertRaises(RuntimeError):
layer.padding = "valid"
with self.assertRaises(RuntimeError):
layer.extra_pad_end = True
with self.assertRaises(RuntimeError):
layer.channel_separable = True
with self.assertRaises(RuntimeError):
layer.data_format = "channels_first"
with self.assertRaises(RuntimeError):
layer.activation = tf.nn.relu
with self.assertRaises(RuntimeError):
layer.use_bias = False
with self.assertRaises(RuntimeError):
layer.use_explicit = False
with self.assertRaises(RuntimeError):
layer.kernel_parameter = tf.ones((1, 2, 3))
with self.assertRaises(RuntimeError):
layer.bias_parameter = tf.ones((3, ))
with self.assertRaises(RuntimeError):
layer.kernel_initializer = tf.keras.initializers.Ones()
with self.assertRaises(RuntimeError):
layer.bias_initializer = tf.keras.initializers.Ones()
with self.assertRaises(RuntimeError):
layer.kernel_regularizer = tf.keras.regularizers.L2()
with self.assertRaises(RuntimeError):
layer.bias_regularizer = tf.keras.regularizers.L2()
def test_invalid_data_format_raises_error(self):
with self.assertRaises(ValueError):
signal_conv.SignalConv1D(2, 1, data_format="NHWC")
