Merge pull request #170 from f0k/remove-args

Remove *args from get_output() and get_output_for()

lasagne/layers/base.py

@@ -91,7 +91,7 @@ def get_output_shape(self):
         """
         return self.get_output_shape_for(self.input_shape)
 
-    def get_output(self, input=None, *args, **kwargs):
+    def get_output(self, input=None, **kwargs):
         """
         Computes the output of the network at this layer. Optionally, you can
         define an input to propagate through the network instead of using the
@@ -125,8 +125,8 @@ def get_output(self, input=None, *args, **kwargs):
                                "there isn't anything to get its input from. "
                                "Did you mean get_output_for()?")
         else:  # in all other cases, just pass the input on to the next layer.
-            layer_input = self.input_layer.get_output(input, *args, **kwargs)
-            return self.get_output_for(layer_input, *args, **kwargs)
+            layer_input = self.input_layer.get_output(input, **kwargs)
+            return self.get_output_for(layer_input, **kwargs)
 
     def get_output_shape_for(self, input_shape):
         """
@@ -153,7 +153,7 @@ def get_output_shape_for(self, input_shape):
         """
         return input_shape
 
-    def get_output_for(self, input, *args, **kwargs):
+    def get_output_for(self, input, **kwargs):
         """
         Propagates the given input through this layer (and only this layer).
 
@@ -272,7 +272,7 @@ def __init__(self, incomings, name=None):
     def get_output_shape(self):
         return self.get_output_shape_for(self.input_shapes)
 
-    def get_output(self, input=None, *args, **kwargs):
+    def get_output(self, input=None, **kwargs):
         if isinstance(input, dict) and (self in input):
             # this layer is mapped to an expression or numpy array
             return utils.as_theano_expression(input[self])
@@ -282,9 +282,9 @@ def get_output(self, input=None, *args, **kwargs):
                                "Did you mean get_output_for()?")
         # In all other cases, just pass the network input on to the next layers
         else:
-            layer_inputs = [input_layer.get_output(input, *args, **kwargs) for
+            layer_inputs = [input_layer.get_output(input, **kwargs) for
                             input_layer in self.input_layers]
-            return self.get_output_for(layer_inputs, *args, **kwargs)
+            return self.get_output_for(layer_inputs, **kwargs)
 
     def get_output_shape_for(self, input_shapes):
         """
@@ -310,7 +310,7 @@ def get_output_shape_for(self, input_shapes):
         """
         raise NotImplementedError
 
-    def get_output_for(self, inputs, *args, **kwargs):
+    def get_output_for(self, inputs, **kwargs):
         """
         Propagates the given inputs through this layer (and only this layer).
 

lasagne/layers/conv.py

@@ -81,7 +81,7 @@ def get_output_shape_for(self, input_shape):
 
         return (input_shape[0], self.num_filters, output_length)
 
-    def get_output_for(self, input, input_shape=None, *args, **kwargs):
+    def get_output_for(self, input, input_shape=None, **kwargs):
         # the optional input_shape argument is for when get_output_for is
         # called directly with a different shape than self.input_shape.
         if input_shape is None:
@@ -171,7 +171,7 @@ def get_output_shape_for(self, input_shape):
 
         return (input_shape[0], self.num_filters, output_rows, output_columns)
 
-    def get_output_for(self, input, input_shape=None, *args, **kwargs):
+    def get_output_for(self, input, input_shape=None, **kwargs):
         # the optional input_shape argument is for when get_output_for is
         # called directly with a different shape than self.input_shape.
         if input_shape is None:

lasagne/layers/corrmm.py

@@ -108,7 +108,7 @@ def get_output_shape_for(self, input_shape):
 
         return (batch_size, self.num_filters, output_rows, output_columns)
 
-    def get_output_for(self, input, *args, **kwargs):
+    def get_output_for(self, input, **kwargs):
         filters = self.W
         if self.flip_filters:
             filters = filters[:, :, ::-1, ::-1]  # flip top-down, left-right

lasagne/layers/cuda_convnet.py

@@ -149,7 +149,7 @@ def get_output_shape_for(self, input_shape):
         else:
             return (self.num_filters, output_rows, output_columns, batch_size)
 
-    def get_output_for(self, input, *args, **kwargs):
+    def get_output_for(self, input, **kwargs):
         if self.dimshuffle:
             filters = self.W.dimshuffle(1, 2, 3, 0)  # bc01 to c01b
             input = input.dimshuffle(1, 2, 3, 0)  # bc01 to c01b
@@ -243,7 +243,7 @@ def get_output_shape_for(self, input_shape):
             return (num_input_channels, output_rows, output_columns,
                     batch_size)
 
-    def get_output_for(self, input, *args, **kwargs):
+    def get_output_for(self, input, **kwargs):
         if self.dimshuffle:
             input = input.dimshuffle(1, 2, 3, 0)  # bc01 to c01b
 
@@ -271,7 +271,7 @@ class ShuffleBC01ToC01BLayer(Layer):
     def get_output_shape_for(self, input_shape):
         return (input_shape[1], input_shape[2], input_shape[3], input_shape[0])
 
-    def get_output_for(self, input, *args, **kwargs):
+    def get_output_for(self, input, **kwargs):
         return input.dimshuffle(1, 2, 3, 0)
 
 bc01_to_c01b = ShuffleBC01ToC01BLayer  # shortcut
@@ -286,7 +286,7 @@ class ShuffleC01BToBC01Layer(Layer):
     def get_output_shape_for(self, input_shape):
         return (input_shape[3], input_shape[0], input_shape[1], input_shape[2])
 
-    def get_output_for(self, input, *args, **kwargs):
+    def get_output_for(self, input, **kwargs):
         return input.dimshuffle(3, 0, 1, 2)
 
 c01b_to_bc01 = ShuffleC01BToBC01Layer  # shortcut
@@ -332,7 +332,7 @@ def get_bias_params(self):
     def get_output_shape_for(self, input_shape):
         return (self.num_units,) + input_shape[1:]
 
-    def get_output_for(self, input, *args, **kwargs):
+    def get_output_for(self, input, **kwargs):
         # fc * c01b... = f01b...
         out = T.tensordot(self.W, input, axes=[[1], [0]])
 

lasagne/layers/dense.py

@@ -77,7 +77,7 @@ def get_bias_params(self):
     def get_output_shape_for(self, input_shape):
         return (input_shape[0], self.num_units)
 
-    def get_output_for(self, input, *args, **kwargs):
+    def get_output_for(self, input, **kwargs):
         if input.ndim > 2:
             # if the input has more than two dimensions, flatten it into a
             # batch of feature vectors.
@@ -131,7 +131,7 @@ def get_bias_params(self):
     def get_output_shape_for(self, input_shape):
         return (input_shape[0], self.num_units) + input_shape[2:]
 
-    def get_output_for(self, input, *args, **kwargs):
+    def get_output_for(self, input, **kwargs):
         # cf * bc01... = fb01...
         out_r = T.tensordot(self.W, input, axes=[[0], [1]])
         # input dims to broadcast over

lasagne/layers/dnn.py

@@ -36,7 +36,7 @@ def get_output_shape_for(self, input_shape):
         output_shape[3] = (output_shape[3] - self.ds[1]) // self.strides[1] + 1
         return tuple(output_shape)
 
-    def get_output_for(self, input, *args, **kwargs):
+    def get_output_for(self, input, **kwargs):
         return dnn.dnn_pool(input, self.ds, self.strides, self.mode)
 
 
@@ -133,7 +133,7 @@ def get_output_shape_for(self, input_shape):
 
         return (batch_size, self.num_filters, output_rows, output_columns)
 
-    def get_output_for(self, input, *args, **kwargs):
+    def get_output_for(self, input, **kwargs):
         # by default we assume 'cross', consistent with corrmm.
         conv_mode = 'conv' if self.flip_filters else 'cross'
         # if 'border_mode' is one of 'valid' or 'full' use that.

lasagne/layers/input.py

@@ -56,7 +56,7 @@ def __init__(self, shape, input_var=None, name=None, **kwargs):
     def get_output_shape(self):
         return self.shape
 
-    def get_output(self, input=None, *args, **kwargs):
+    def get_output(self, input=None, **kwargs):
         if isinstance(input, dict):
             input = input.get(self, None)
         if input is None:

lasagne/layers/merge.py

@@ -21,7 +21,7 @@ def get_output_shape_for(self, input_shapes):
         output_shape[self.axis] = sum(sizes)
         return tuple(output_shape)
 
-    def get_output_for(self, inputs, *args, **kwargs):
+    def get_output_for(self, inputs, **kwargs):
         return T.concatenate(inputs, axis=self.axis)
 
 concat = ConcatLayer  # shortcut
@@ -66,7 +66,7 @@ def get_output_shape_for(self, input_shapes):
             raise ValueError("Mismatch: not all input shapes are the same")
         return input_shapes[0]
 
-    def get_output_for(self, inputs, *args, **kwargs):
+    def get_output_for(self, inputs, **kwargs):
         output = None
         for coeff, input in zip(self.coeffs, inputs):
             if coeff != 1:

lasagne/layers/noise.py

@@ -20,7 +20,7 @@ def __init__(self, incoming, p=0.5, rescale=True, **kwargs):
         self.p = p
         self.rescale = rescale
 
-    def get_output_for(self, input, deterministic=False, *args, **kwargs):
+    def get_output_for(self, input, deterministic=False, **kwargs):
         if deterministic or self.p == 0:
             return input
         else:
@@ -44,7 +44,7 @@ def __init__(self, incoming, sigma=0.1, **kwargs):
         super(GaussianNoiseLayer, self).__init__(incoming, **kwargs)
         self.sigma = sigma
 
-    def get_output_for(self, input, deterministic=False, *args, **kwargs):
+    def get_output_for(self, input, deterministic=False, **kwargs):
         if deterministic or self.sigma == 0:
             return input
         else:

lasagne/layers/normalization.py

@@ -86,7 +86,7 @@ def __init__(self, incoming, alpha=1e-4, k=2, beta=0.75, n=5, **kwargs):
     def get_output_shape_for(self, input_shape):
         return input_shape
 
-    def get_output_for(self, input, *args, **kwargs):
+    def get_output_for(self, input, **kwargs):
         input_shape = self.input_shape
         if any(s is None for s in input_shape):
             input_shape = input.shape

lasagne/layers/pool.py

@@ -36,7 +36,7 @@ def get_output_shape_for(self, input_shape):
 
         return tuple(output_shape)
 
-    def get_output_for(self, input, *args, **kwargs):
+    def get_output_for(self, input, **kwargs):
         return downsample.max_pool_2d(input, self.ds, self.ignore_border)
 
 
@@ -74,7 +74,7 @@ def get_output_shape_for(self, input_shape):
         output_shape[self.axis] = output_shape[self.axis] // self.ds
         return tuple(output_shape)
 
-    def get_output_for(self, input, *args, **kwargs):
+    def get_output_for(self, input, **kwargs):
         num_feature_maps = input.shape[self.axis]
         num_feature_maps_out = num_feature_maps // self.ds
 
@@ -113,7 +113,7 @@ def __init__(self, incoming, ds, axis=1, **kwargs):
                                "multiple of the group size (ds=%d)" %
                                (num_feature_maps, self.ds))
 
-    def get_output_for(self, input, *args, **kwargs):
+    def get_output_for(self, input, **kwargs):
         num_feature_maps = input.shape[self.axis]
         num_pools = num_feature_maps // self.ds
 
@@ -151,5 +151,5 @@ def __init__(self, incoming, pool_function=T.mean, **kwargs):
     def get_output_shape_for(self, input_shape):
         return input_shape[:2]
 
-    def get_output_for(self, input, *args, **kwargs):
+    def get_output_for(self, input, **kwargs):
         return self.pool_function(input.flatten(3), axis=2)

lasagne/layers/shape.py

@@ -21,7 +21,7 @@ class FlattenLayer(Layer):
     def get_output_shape_for(self, input_shape):
         return (input_shape[0], int(np.prod(input_shape[1:])))
 
-    def get_output_for(self, input, *args, **kwargs):
+    def get_output_for(self, input, **kwargs):
         return input.flatten(2)
 
 flatten = FlattenLayer  # shortcut
@@ -80,7 +80,7 @@ def __init__(self, incoming, shape, **kwargs):
             raise ValueError("`shape` cannot contain multiple -1")
         self.shape = shape
 
-    def get_output_shape_for(self, input_shape, *args, **kwargs):
+    def get_output_shape_for(self, input_shape, **kwargs):
         # Initialize output shape from shape specification
         output_shape = list(self.shape)
         # First, replace all `[i]` with the corresponding input dimension, and
@@ -127,7 +127,7 @@ def get_output_shape_for(self, input_shape, *args, **kwargs):
                              (input_shape, self.shape))
         return tuple(output_shape)
 
-    def get_output_for(self, input, *args, **kwargs):
+    def get_output_for(self, input, **kwargs):
         # Replace all `[i]` with the corresponding input dimension
         output_shape = list(self.shape)
         for dim, o in enumerate(output_shape):
@@ -226,7 +226,7 @@ def get_output_shape_for(self, input_shape):
 
         return tuple(output_shape)
 
-    def get_output_for(self, input, *args, **kwargs):
+    def get_output_for(self, input, **kwargs):
         return input.dimshuffle(self.pattern)
 
 dimshuffle = DimshuffleLayer  # shortcut
@@ -249,7 +249,7 @@ def get_output_shape_for(self, input_shape):
 
         return output_shape
 
-    def get_output_for(self, input, *args, **kwargs):
+    def get_output_for(self, input, **kwargs):
         return padding.pad(input, self.width, self.val, self.batch_ndim)
 
 pad = PadLayer  # shortcut

lasagne/objectives.py

@@ -50,8 +50,7 @@ def __init__(self, input_layer, loss_function=mse, aggregation='mean'):
                              'or None, not {0}'.format(aggregation))
         self.aggregation = aggregation
 
-    def get_loss(self, input=None, target=None, aggregation=None,
-                 *args, **kwargs):
+    def get_loss(self, input=None, target=None, aggregation=None, **kwargs):
         """
         Get loss scalar expression
 
@@ -63,15 +62,13 @@ def get_loss(self, input=None, target=None, aggregation=None,
                 given the input
             - aggregation : None to use the value passed to the
                 constructor or a value to override it
-            - args : additional arguments passed to `input_layer`'s
-                `get_output` method
             - kwargs : additional keyword arguments passed to `input_layer`'s
                 `get_output` method
 
         :returns:
             - output : loss expressions
         """
-        network_output = self.input_layer.get_output(input, *args, **kwargs)
+        network_output = self.input_layer.get_output(input, **kwargs)
         if target is None:
             target = self.target_var
         if aggregation not in self._valid_aggregation:
@@ -130,7 +127,7 @@ def __init__(self, input_layer, loss_function=mse, aggregation='mean'):
         self.aggregation = aggregation
 
     def get_loss(self, input=None, target=None, mask=None,
-                 aggregation=None, *args, **kwargs):
+                 aggregation=None, **kwargs):
         """
         Get loss scalar expression
 
@@ -146,15 +143,13 @@ def get_loss(self, input=None, target=None, mask=None,
                 contributions of the resulting loss values
             - aggregation : None to use the value passed to the
                 constructor or a value to override it
-            - args : additional arguments passed to `input_layer`'s
-                `get_output` method
             - kwargs : additional keyword arguments passed to `input_layer`'s
                 `get_output` method
 
         :returns:
             - output : loss expressions
         """
-        network_output = self.input_layer.get_output(input, *args, **kwargs)
+        network_output = self.input_layer.get_output(input, **kwargs)
         if target is None:
             target = self.target_var
         if mask is None:

lasagne/tests/layers/test_base.py

@@ -26,15 +26,15 @@ def test_get_output_without_arguments(self, layer):
         layer.input_layer.get_output.assert_called_with(None)
 
     def test_get_output_passes_on_arguments_to_input_layer(self, layer):
-        input, arg, kwarg = object(), object(), object()
+        input, kwarg = object(), object()
         layer.get_output_for = Mock()
 
-        output = layer.get_output(input, arg, kwarg=kwarg)
+        output = layer.get_output(input, kwarg=kwarg)
         assert output is layer.get_output_for.return_value
         layer.get_output_for.assert_called_with(
-            layer.input_layer.get_output.return_value, arg, kwarg=kwarg)
+            layer.input_layer.get_output.return_value, kwarg=kwarg)
         layer.input_layer.get_output.assert_called_with(
-            input, arg, kwarg=kwarg)
+            input, kwarg=kwarg)
 
     def test_get_output_input_is_a_mapping(self, layer):
         input = {layer: theano.tensor.matrix()}
@@ -140,19 +140,19 @@ def test_get_output_without_arguments(self, layer):
         layer.input_layers[1].get_output.assert_called_with(None)
 
     def test_get_output_passes_on_arguments_to_input_layer(self, layer):
-        input, arg, kwarg = object(), object(), object()
+        input, kwarg = object(), object()
         layer.get_output_for = Mock()
 
-        output = layer.get_output(input, arg, kwarg=kwarg)
+        output = layer.get_output(input, kwarg=kwarg)
         assert output is layer.get_output_for.return_value
         layer.get_output_for.assert_called_with([
             layer.input_layers[0].get_output.return_value,
             layer.input_layers[1].get_output.return_value,
-            ], arg, kwarg=kwarg)
+            ], kwarg=kwarg)
         layer.input_layers[0].get_output.assert_called_with(
-            input, arg, kwarg=kwarg)
+            input, kwarg=kwarg)
         layer.input_layers[1].get_output.assert_called_with(
-            input, arg, kwarg=kwarg)
+            input, kwarg=kwarg)
 
     def test_get_output_input_is_a_mapping(self, layer):
         input = {layer: theano.tensor.matrix()}