Merge pull request #35 from philipperemy/issue_34

Small updates

Author: Philippe Rémy

README.md

@@ -92,7 +92,7 @@ model.fit(x, y) # Keras model.
 
 ### Arguments
 
-`tcn.TCN(nb_filters=64, kernel_size=2, nb_stacks=1, dilations=None, activation='norm_relu', padding='causal', use_skip_connections=True, dropout_rate=0.0, return_sequences=True, name='tcn')`
+`tcn.TCN(nb_filters=64, kernel_size=2, nb_stacks=1, dilations=[1, 2, 4, 8, 16, 32], activation='norm_relu', padding='causal', use_skip_connections=True, dropout_rate=0.0, return_sequences=True, name='tcn')`
 
 - `nb_filters`: Integer. The number of filters to use in the convolutional layers.
 - `kernel_size`: Integer. The size of the kernel to use in each convolutional layer.

setup.py

@@ -2,7 +2,7 @@
 
 setup(
     name='keras-tcn',
-    version='2.3.4',
+    version='2.3.5',
     description='Keras TCN',
     author='Philippe Remy',
     license='MIT',

tcn/tcn.py

@@ -45,7 +45,7 @@ def wave_net_activation(x):
 
 
 def residual_block(x, s, i, activation, nb_filters, kernel_size, padding, dropout_rate=0, name=''):
-    # type: (Layer, int, int, str, int, int, float, str) -> Tuple[Layer, Layer]
+    # type: (Layer, int, int, str, int, int, str, float, str) -> Tuple[Layer, Layer]
     """Defines the residual block for the WaveNet TCN
 
     Args:
@@ -67,7 +67,7 @@ def residual_block(x, s, i, activation, nb_filters, kernel_size, padding, dropou
     original_x = x
     conv = Conv1D(filters=nb_filters, kernel_size=kernel_size,
                   dilation_rate=i, padding=padding,
-                  name=name + '_dilated_conv_%d_tanh_s%d' % (i, s))(x)
+                  name=name + '_d_%s_conv_%d_tanh_s%d' % (padding, i, s))(x)
     if activation == 'norm_relu':
         x = Activation('relu')(conv)
         x = Lambda(channel_normalization)(x)
@@ -100,8 +100,10 @@ def is_power_of_two(num):
 class TCN:
     """Creates a TCN layer.
 
+        Input shape:
+            A tensor of shape (batch_size, timesteps, input_dim).
+
         Args:
-            input_layer: A tensor of shape (batch_size, timesteps, input_dim).
             nb_filters: The number of filters to use in the convolutional layers.
             kernel_size: The size of the kernel to use in each convolutional layer.
             dilations: The list of the dilations. Example is: [1, 2, 4, 8, 16, 32, 64].
@@ -121,7 +123,7 @@ def __init__(self,
                  nb_filters=64,
                  kernel_size=2,
                  nb_stacks=1,
-                 dilations=None,
+                 dilations=[1, 2, 4, 8, 16, 32],
                  activation='norm_relu',
                  padding='causal',
                  use_skip_connections=True,
@@ -139,12 +141,8 @@ def __init__(self,
         self.nb_filters = nb_filters
         self.padding = padding
 
-        # backwards incompatibility warning.
-        # o = tcn.TCN(i, return_sequences=False) =>
-        # o = tcn.TCN(return_sequences=False)(i)
-
         if padding != 'causal' and padding != 'same':
-            raise ValueError("Only 'causal' or 'same' paddings are compatible for this layer.")
+            raise ValueError("Only 'causal' or 'same' padding are compatible for this layer.")
 
         if not isinstance(nb_filters, int):
             print('An interface change occurred after the version 2.1.2.')
@@ -154,8 +152,6 @@ def __init__(self,
             raise Exception()
 
     def __call__(self, inputs):
-        if self.dilations is None:
-            self.dilations = [1, 2, 4, 8, 16, 32]
         x = inputs
         x = Convolution1D(self.nb_filters, 1, padding=self.padding, name=self.name + '_initial_conv')(x)
         skip_connections = []