Resolve "custom dense layers in rnn"

Closes #361 (closed)

mlair/model_modules/recurrent_networks.py

@@ -131,7 +131,7 @@ class RNN(AbstractModelClass):  # pragma: no cover
         if self.add_dense_layer is True:
             if len(self.dense_layer_configuration) == 0:
                 x_in = keras.layers.Dense(min(self._output_shape ** 2, conf[-1]), name=f"Dense_{len(conf) + 1}",
-                                        kernel_initializer=self.kernel_initializer, )(x_in)
+                                          kernel_initializer=self.kernel_initializer, )(x_in)
                 x_in = self.activation(name=f"{self.activation_name}_{len(conf) + 1}")(x_in)
                 if self.dropout is not None:
                     x_in = self.dropout(self.dropout_rate)(x_in)
@@ -139,8 +139,8 @@ class RNN(AbstractModelClass):  # pragma: no cover
                 for layer, n_hidden in enumerate(self.dense_layer_configuration):
                     if n_hidden < self._output_shape:
                         break
-                    x_in = keras.layers.Dense(n_hidden), name=f"Dense_{layer + 1}",
-                                            kernel_initializer=self.kernel_initializer, )(x_in)
+                    x_in = keras.layers.Dense(n_hidden, name=f"Dense_{layer + 1}",
+                                              kernel_initializer=self.kernel_initializer, )(x_in)
                     x_in = self.activation(name=f"{self.activation_name}_{layer + 1}")(x_in)
                     if self.dropout is not None:
                         x_in = self.dropout(self.dropout_rate)(x_in)