Replace __call__ with forward

chainer/link.py

@@ -108,11 +108,11 @@ def __init__(self, n_in, n_out):
                       self.b = chainer.Parameter(
                           initializers.Zero(), (n_out,))
 
-              def __call__(self, x):
+              def forward(self, x):
                   return F.linear(x, self.W, self.b)
 
        This example shows that a user can define arbitrary parameters and use
-       them in any methods. Links typically implement the ``__call__``
+       them in any methods. Links typically implement the ``forward``
        operator, although they can also provide other methods to implement the
        forward propagation.
 
@@ -194,14 +194,8 @@ def __init__(self):
         finally:
             self._within_init_scope = old_flag
 
-    def __call__(self, *args, **kwargs):
-        try:
-            forward = self.forward
-        except AttributeError:
-            raise TypeError(
-                '{} object has neither \'Link.__call__\' method overridden'
-                ' nor \'forward\' method defined.'.format(self))
-        return forward(*args, **kwargs)
+    def forward(self, *args, **kwargs):
+        return self.forward(*args, **kwargs)
 
     def __setattr__(self, name, value):
         if self.within_init_scope and isinstance(value, variable.Parameter):
@@ -643,7 +637,7 @@ def __init__(self):
                                 None, 64, 3, 1, 1, nobias=True)
                             self.bn = L.BatchNormalization(64)
 
-                    def __call__(self, x):
+                    def forward(self, x):
                         return F.relu(self.bn(self.conv(x)))
 
                 net = ConvBNReLU().repeat(16, mode='init')
@@ -772,15 +766,15 @@ def __init__(self, n_in, n_hidden, n_out):
                       self.layer2 = L.Linear(n_hidden, n_hidden)
                       self.layer3 = L.Linear(n_hidden, n_out)
 
-              def __call__(self, x):
+              def forward(self, x):
                   # Forward propagation
                   h1 = F.relu(self.layer1(x))
                   h2 = F.relu(self.layer2(h1))
                   return self.layer3(h2)
 
        Child links are registered via the assignment within a
        ``with self.init_scope():`` block. The forward propagation is often
-       implemented as the ``__call__`` operator as the above example, though
+       implemented as the ``forward`` operator as the above example, though
        it is not mandatory.
 
     Args:

chainer/links/activation/maxout.py

@@ -80,7 +80,7 @@ def __init__(self, in_size, out_size, pool_size,
         self.out_size = out_size
         self.pool_size = pool_size
 
-    def __call__(self, x):
+    def forward(self, x):
         """Applies the maxout layer.
 
         Args:

chainer/links/activation/prelu.py

@@ -27,7 +27,7 @@ def __init__(self, shape=(), init=0.25):
         with self.init_scope():
             self.W = variable.Parameter(init, shape)
 
-    def __call__(self, x):
+    def forward(self, x):
         """Applies the parametric ReLU activation function.
 
         Args:

chainer/links/activation/simplified_dropconnect.py

@@ -71,7 +71,7 @@ def __init__(self, in_size, out_size, ratio=.5, nobias=False,
     def _initialize_params(self, in_size):
         self.W.initialize((self.out_size, in_size))
 
-    def __call__(self, x, train=True, mask=None, use_batchwise_mask=True):
+    def forward(self, x, train=True, mask=None, use_batchwise_mask=True):
         """Applies the simplified dropconnect layer.
 
         Args:

chainer/links/activation/swish.py

@@ -35,7 +35,7 @@ def __init__(self, n_units, n_out):
                     self.l2 = L.Linear(None, n_units)
                     self.l3 = L.Linear(None, n_out)
 
-            def __call__(self, x):
+            def forward(self, x):
                 h1 = F.relu(self.l1(x))
                 h2 = F.relu(self.l2(h1))
                 return self.l3(h2)
@@ -53,7 +53,7 @@ def __init__(self, n_units, n_out):
                     self.s2 = L.Swish(None)
                     self.l3 = L.Linear(None, n_out)
 
-            def __call__(self, x):
+            def forward(self, x):
                 h1 = self.s1(self.l1(x))
                 h2 = self.s2(self.l2(h1))
                 return self.l3(h2)
@@ -77,7 +77,7 @@ def __init__(self, beta_shape, beta_init=1.0):
                 beta_init = initializers.Constant(beta_init)
                 self.beta = variable.Parameter(beta_init)
 
-    def __call__(self, x):
+    def forward(self, x):
         """Applies the Swish activation function.
 
         Args:

chainer/links/caffe/caffe_function.py

@@ -173,8 +173,8 @@ def __init__(self, model_path):
                         'Skip the layer "%s", since CaffeFunction does not'
                         'support it' % layer.name)
 
-    def __call__(self, inputs, outputs, disable=(), **kwargs):
-        """__call__(self, inputs, outputs, disable=())
+    def forward(self, inputs, outputs, disable=(), **kwargs):
+        """forward(self, inputs, outputs, disable=())
 
         Executes a sub-network of the network.
 

chainer/links/connection/bias.py

@@ -17,7 +17,7 @@ class Bias(link.Link):
             input is applied.
         shape (tuple of ints): Shape of the learnable bias parameter. If
             ``None``, this link does not have learnable parameters so an
-            explicit bias needs to be given to its ``__call__`` method's second
+            explicit bias needs to be given to its ``forward`` method's second
             input.
 
     .. seealso:: See :func:`~chainer.functions.bias` for details.
@@ -38,7 +38,7 @@ def __init__(self, axis=1, shape=None):
 
         self.axis = axis
 
-    def __call__(self, *xs):
+    def forward(self, *xs):
         """Applies broadcasted elementwise summation.
 
         Args:

chainer/links/connection/bilinear.py

@@ -89,7 +89,7 @@ def __init__(self, left_size, right_size, out_size, nobias=False,
                 self.V2 = variable.Parameter(initialV2, V2_shape)
                 self.b = variable.Parameter(initialb, b_shape)
 
-    def __call__(self, e1, e2):
+    def forward(self, e1, e2):
         """Applies the bilinear function to inputs and the internal parameters.
 
         Args:

chainer/links/connection/convolution_2d.py

@@ -156,7 +156,7 @@ def _initialize_params(self, in_channels):
         W_shape = (self.out_channels, int(in_channels / self.groups), kh, kw)
         self.W.initialize(W_shape)
 
-    def __call__(self, x):
+    def forward(self, x):
         """Applies the convolution layer.
 
         Args:

chainer/links/connection/convolution_nd.py

@@ -82,7 +82,7 @@ def __init__(self, ndim, in_channels, out_channels, ksize, stride=1, pad=0,
                 initial_bias = initializers._get_initializer(initial_bias)
                 self.b = variable.Parameter(initial_bias, out_channels)
 
-    def __call__(self, x):
+    def forward(self, x):
         """Applies N-dimensional convolution layer.
 
         Args:

chainer/links/connection/deconvolution_2d.py

@@ -175,7 +175,7 @@ def _initialize_params(self, in_channels):
         W_shape = (in_channels, int(self.out_channels / self.groups), kh, kw)
         self.W.initialize(W_shape)
 
-    def __call__(self, x):
+    def forward(self, x):
         if self.W.data is None:
             self._initialize_params(x.shape[1])
         return deconvolution_2d.deconvolution_2d(

chainer/links/connection/deconvolution_nd.py

@@ -79,7 +79,7 @@ def __init__(self, ndim, in_channels, out_channels, ksize, stride=1, pad=0,
                 initial_bias = initializers._get_initializer(initial_bias)
                 self.b = variable.Parameter(initial_bias, out_channels)
 
-    def __call__(self, x):
+    def forward(self, x):
         return deconvolution_nd.deconvolution_nd(
             x, self.W, b=self.b, stride=self.stride, pad=self.pad,
             outsize=self.outsize, dilate=self.dilate, groups=self.groups)

chainer/links/connection/deformable_convolution_2d.py

@@ -69,7 +69,7 @@ def __init__(self, in_channels, out_channels, ksize, stride=1, pad=0,
                 in_channels, out_channels, ksize, stride, pad,
                 deform_nobias, deform_initialW, deform_initial_bias)
 
-    def __call__(self, x):
+    def forward(self, x):
         """Applies the deformable convolution.
 
         Args:
@@ -121,7 +121,7 @@ def _initialize_params(self, in_channels):
         if self.b is not None:
             self.b.initialize(self.out_channels)
 
-    def __call__(self, x, offset):
+    def forward(self, x, offset):
         if self.W.data is None:
             self._initialize_params(x.shape[1])
         return deformable_convolution_2d_sampler(

chainer/links/connection/depthwise_convolution_2d.py

@@ -76,7 +76,7 @@ def _initialize_params(self, in_channels):
         if self.b is not None:
             self.b.initialize(self.channel_multiplier * in_channels)
 
-    def __call__(self, x):
+    def forward(self, x):
         """Applies the depthwise convolution layer.
 
         Args:

chainer/links/connection/dilated_convolution_2d.py

@@ -118,7 +118,7 @@ def _initialize_params(self, in_channels):
         W_shape = (self.out_channels, in_channels, kh, kw)
         self.W.initialize(W_shape)
 
-    def __call__(self, x):
+    def forward(self, x):
         """Applies the convolution layer.
 
         Args:

chainer/links/connection/embed_id.py

@@ -57,7 +57,7 @@ def __init__(self, in_size, out_size, initialW=None, ignore_label=None):
                 initialW = normal.Normal(1.0)
             self.W = variable.Parameter(initialW, (in_size, out_size))
 
-    def __call__(self, x):
+    def forward(self, x):
         """Extracts the word embedding of given IDs.
 
         Args:

chainer/links/connection/gru.py

@@ -105,7 +105,7 @@ class StatelessGRU(GRUBase):
 
     """
 
-    def __call__(self, h, x):
+    def forward(self, h, x):
         r = sigmoid.sigmoid(self.W_r(x) + self.U_r(h))
         z = sigmoid.sigmoid(self.W_z(x) + self.U_z(h))
         h_bar = tanh.tanh(self.W(x) + self.U(r * h))
@@ -223,7 +223,7 @@ def set_state(self, h):
     def reset_state(self):
         self.h = None
 
-    def __call__(self, x):
+    def forward(self, x):
         z = self.W_z(x)
         h_bar = self.W(x)
         if self.h is not None:
@@ -261,15 +261,15 @@ class GRU(StatefulGRU):
 
     """
 
-    def __call__(self, *args):
-        """__call__(self, x)
+    def forward(self, *args):
+        """forward(self, x)
 
         Does forward propagation.
 
         """
 
         n_args = len(args)
-        msg = ("Invalid argument. The length of GRU.__call__ must be 1. "
+        msg = ("Invalid argument. The length of GRU.forward must be 1. "
                "But %d is given. " % n_args)
 
         if n_args == 0 or n_args >= 3:
@@ -281,4 +281,4 @@ def __call__(self, *args):
                     "Use chainer.links.StatelessGRU instead.")
             raise ValueError(msg)
 
-        return super(GRU, self).__call__(args[0])
+        return super(GRU, self).forward(args[0])

chainer/links/connection/highway.py

@@ -60,7 +60,7 @@ def __init__(self, in_out_size, nobias=False, activate=relu.relu,
                 in_out_size, in_out_size, nobias=nobias,
                 initialW=init_Wt, initial_bias=init_bt)
 
-    def __call__(self, x):
+    def forward(self, x):
         """Computes the output of the Highway module.
 
         Args:

chainer/links/connection/inception.py

@@ -61,7 +61,7 @@ def __init__(self, in_channels, out1, proj3, out3, proj5, out5, proj_pool,
                 in_channels, proj_pool, 1, initialW=conv_init,
                 initial_bias=bias_init)
 
-    def __call__(self, x):
+    def forward(self, x):
         """Computes the output of the Inception module.
 
         Args:

chainer/links/connection/inceptionbn.py

@@ -92,7 +92,7 @@ def __init__(self, in_channels, out1, proj3, out3, proj33, out33,
                 self.poolpn = batch_normalization.BatchNormalization(
                     proj_pool, dtype=dtype)
 
-    def __call__(self, x):
+    def forward(self, x):
         outs = []
 
         if self.out1 > 0:

chainer/links/connection/linear.py

@@ -113,7 +113,7 @@ def __init__(self, in_size, out_size=None, nobias=False,
     def _initialize_params(self, in_size):
         self.W.initialize((self.out_size, in_size))
 
-    def __call__(self, x):
+    def forward(self, x):
         """Applies the linear layer.
 
         Args:

chainer/links/connection/local_convolution_2d.py

@@ -87,7 +87,7 @@ def _initialize_params(self, in_channels, in_size):
         if not self.nobias:
             self.b.initialize(bias_shape)
 
-    def __call__(self, x):
+    def forward(self, x):
         """Applies the local convolution layer.
 
         Args:

chainer/links/connection/lstm.py

@@ -117,7 +117,7 @@ class StatelessLSTM(LSTMBase):
 
     """
 
-    def __call__(self, c, h, x):
+    def forward(self, c, h, x):
         """Returns new cell state and updated output of LSTM.
 
         Args:
@@ -289,7 +289,7 @@ def reset_state(self):
         """
         self.c = self.h = None
 
-    def __call__(self, x):
+    def forward(self, x):
         """Updates the internal state and returns the LSTM outputs.
 
         Args:

chainer/links/connection/mgu.py

@@ -26,7 +26,7 @@ def _call_mgu(self, h, x):
 
 class StatelessMGU(MGUBase):
 
-    __call__ = MGUBase._call_mgu
+    forward = MGUBase._call_mgu
 
 
 class StatefulMGU(MGUBase):
@@ -58,7 +58,7 @@ def set_state(self, h):
     def reset_state(self):
         self.h = None
 
-    def __call__(self, x):
+    def forward(self, x):
         if self.h is None:
             n_batch = x.shape[0]
             h_data = self.xp.zeros(

chainer/links/connection/mlp_convolution_2d.py

@@ -84,7 +84,7 @@ def __init__(self, in_channels, out_channels, ksize=None, stride=1, pad=0,
         super(MLPConvolution2D, self).__init__(*convs)
         self.activation = activation
 
-    def __call__(self, x):
+    def forward(self, x):
         """Computes the output of the mlpconv layer.
 
         Args:

chainer/links/connection/n_step_lstm.py

@@ -31,8 +31,8 @@ class NStepLSTMBase(n_step_rnn.NStepRNNBase):
 
     n_weights = 8
 
-    def __call__(self, hx, cx, xs, **kwargs):
-        """__call__(self, hx, cx, xs)
+    def forward(self, hx, cx, xs, **kwargs):
+        """forward(self, hx, cx, xs)
 
         Calculate all hidden states and cell states.
 

chainer/links/connection/n_step_rnn.py

@@ -126,8 +126,8 @@ def n_cells(self):
         """
         return NotImplementedError
 
-    def __call__(self, hx, xs, **kwargs):
-        """__call__(self, hx, xs)
+    def forward(self, hx, xs, **kwargs):
+        """forward(self, hx, xs)
 
         Calculate all hidden states and cell states.
 

chainer/links/connection/parameter.py

@@ -26,7 +26,7 @@ def __init__(self, array):
         if isinstance(array, cuda.ndarray):
             self.to_gpu(cuda.get_device_from_array(array))
 
-    def __call__(self, volatile='off'):
+    def forward(self, volatile='off'):
         """Returns the parameter variable.
 
         Args: