test=develop, hide build_once (#17871)

python/paddle/fluid/dygraph/layers.py

@@ -150,12 +150,12 @@ class Layer(core.Layer):
             if p.trainable:
                 p.clear_gradient()
 
-    def build_once(self, *args):
+    def _build_once(self, *args):
         pass
 
     def __call__(self, *inputs):
         if not self._built:
-            self.build_once(*inputs)
+            self._build_once(*inputs)
             if parallel_helper._is_data_parallel_mode():
                 parallel_helper._broadcast_parameters(self._parameters.values())
 

python/paddle/fluid/dygraph/nn.py

@@ -372,7 +372,7 @@ class Conv3D(layers.Layer):
         self._param_attr = param_attr
         self._bias_attr = bias_attr
 
-    def build_once(self, input):
+    def _build_once(self, input):
         num_channels = input.shape[1]
         self._dtype = self._helper.input_dtype(input)
 
@@ -579,7 +579,7 @@ class Conv3DTranspose(layers.Layer):
         self._bias_attr = bias_attr
         self._act = act
 
-    def build_once(self, input):
+    def _build_once(self, input):
         self._dtype = self._helper.input_dtype(input)
         self._input_channel = input.shape[1]
 
@@ -883,7 +883,7 @@ class FC(layers.Layer):
         assert isinstance(value, Parameter)
         self.__w[i] = value
 
-    def build_once(self, input):
+    def _build_once(self, input):
         i = 0
         for inp, param in self._helper.iter_inputs_and_params(input,
                                                               self._param_attr):
@@ -1112,7 +1112,7 @@ class BatchNorm(layers.Layer):
         self._fuse_with_relu = fuse_with_relu
         self._use_global_stats = use_global_stats
 
-    def build_once(self, input):
+    def _build_once(self, input):
         pass
 
     def forward(self, input):
@@ -1189,6 +1189,7 @@ class Embedding(layers.Layer):
                   supplied inputs.
 
     Examples:
+
         .. code-block:: python
 
           dict_size = len(dataset.ids)
@@ -1316,7 +1317,7 @@ class LayerNorm(layers.Layer):
         self._bias_attr = bias_attr
         self._act = act
 
-    def build_once(self, input):
+    def _build_once(self, input):
         self._dtype = self._helper.input_dtype(input)
         input_shape = input.shape
         param_shape = [
@@ -1678,7 +1679,7 @@ class NCE(layers.Layer):
             'remote_prefetch': remote_prefetch
         }
 
-    def build_once(self, input, label, sample_weight=None):
+    def _build_once(self, input, label, sample_weight=None):
         assert isinstance(input, Variable)
         assert isinstance(label, Variable)
 
@@ -1764,7 +1765,7 @@ class PRelu(layers.Layer):
             raise ValueError('mode should be one of all, channel, element.')
         self._alpha_shape = [1]
 
-    def build_once(self, input):
+    def _build_once(self, input):
         if self._mode == 'channel':
             self._alpha_shape = [1, input.shape[1], 1, 1]
         elif self._mode == 'element':
@@ -1842,7 +1843,7 @@ class BilinearTensorProduct(layers.Layer):
         self._name = name
         self._inputs = dict()
 
-    def build_once(self, x, y):
+    def _build_once(self, x, y):
         self._dtype = self._helper.input_dtype(x)
 
         param_shape = [self._size, x.shape[1], y.shape[1]]
@@ -2018,7 +2019,7 @@ class Conv2DTranspose(layers.Layer):
         self._output_size = output_size
         self._op_type = 'conv2d_transpose'
 
-    def build_once(self, input):
+    def _build_once(self, input):
         input_channel = input.shape[1]
         if (input_channel == self._groups and
                 self._num_filters == input_channel and not self._use_cudnn):
@@ -2142,7 +2143,7 @@ class SequenceConv(layers.Layer):
         self._bias_attr = bias_attr
         self._param_attr = param_attr
 
-    def build_once(self, input):
+    def _build_once(self, input):
         self._dtype = self._helper.input_dtype(input)
         filter_shape = [self._filter_size * input.shape[1], self._num_filters]
         self._filter_param = self.create_parameter(
@@ -2179,7 +2180,7 @@ class RowConv(layers.Layer):
         self._param_attr = param_attr
         self._future_context_size = future_context_size
 
-    def build_once(self, input):
+    def _build_once(self, input):
         self._dtype = self._helper.input_dtype(input)
         filter_shape = [self._future_context_size + 1, input.shape[1]]
         self._filter_param = self.create_parameter(
@@ -2242,7 +2243,7 @@ class GroupNorm(layers.Layer):
         if data_layout != 'NCHW':
             raise ValueError("unsupported data layout:" + data_layout)
 
-    def build_once(self, input):
+    def _build_once(self, input):
         self._dtype = self._helper.input_dtype(input)
         param_shape = [input.shape[1]]
         if self._bias_attr:
@@ -2295,7 +2296,7 @@ class SpectralNorm(layers.Layer):
         self._eps = eps
         self._dim = dim
 
-    def build_once(self, weight):
+    def _build_once(self, weight):
         self._dtype = self._helper.input_dtype(weight)
         input_shape = weight.shape
         h = input_shape[self._dim]
@@ -2350,7 +2351,7 @@ class TreeConv(layers.Layer):
         self._bias_attr = bias_attr
         self._param_attr = param_attr
 
-    def build_once(self, nodes_vector, edge_set):
+    def _build_once(self, nodes_vector, edge_set):
         assert isinstance(nodes_vector, Variable)
         assert isinstance(edge_set, Variable)
         self._dtype = self._helper.input_dtype(nodes_vector)

python/paddle/fluid/tests/unittests/test_imperative_basic.py

@@ -67,7 +67,7 @@ class SimpleRNNCell(fluid.Layer):
         self._dtype = core.VarDesc.VarType.FP32
         self.param_attr = param_attr
 
-    def build_once(self, inputs, pre_hidden):
+    def _build_once(self, inputs, pre_hidden):
         i2h_param_shape = [self.step_input_size, self.hidden_size]
         h2h_param_shape = [self.hidden_size, self.hidden_size]
         h2o_param_shape = [self.output_size, self.hidden_size]

python/paddle/fluid/tests/unittests/test_imperative_ocr_attention_model.py

@@ -307,9 +307,6 @@ class SimpleAttention(fluid.dygraph.Layer):
                        act=None,
                        bias_attr=False)
 
-    def _build_once(self, encoder_vec, encoder_proj, decoder_state):
-        pass
-
     def forward(self, encoder_vec, encoder_proj, decoder_state):
 
         decoder_state_fc = self.fc_1(decoder_state)
@@ -358,10 +355,6 @@ class GRUDecoderWithAttention(fluid.dygraph.Layer):
 
         self.decoder_size = decoder_size
 
-    def _build_once(self, target_embedding, encoder_vec, encoder_proj,
-                    decoder_boot):
-        pass
-
     def forward(self, target_embedding, encoder_vec, encoder_proj,
                 decoder_boot):
         res = []
@@ -402,9 +395,6 @@ class OCRAttention(fluid.dygraph.Layer):
         self.gru_decoder_with_attention = GRUDecoderWithAttention(
             self.full_name(), Config.decoder_size, Config.num_classes)
 
-    def _build_once(self, inputs, label_in):
-        pass
-
     def forward(self, inputs, label_in):
         gru_backward, encoded_vector, encoded_proj = self.encoder_net(inputs)
         backward_first = fluid.layers.slice(

python/paddle/fluid/tests/unittests/test_imperative_ptb_rnn.py

@@ -44,7 +44,7 @@ class SimpleLSTMRNN(fluid.Layer):
         self.cell_array = []
         self.hidden_array = []
 
-    def build_once(self, input_embedding, init_hidden=None, init_cell=None):
+    def _build_once(self, input_embedding, init_hidden=None, init_cell=None):
         self.weight_1_arr = []
         self.weight_2_arr = []
         self.bias_arr = []
@@ -176,9 +176,6 @@ class PtbModel(fluid.Layer):
             default_initializer=fluid.initializer.UniformInitializer(
                 low=-self.init_scale, high=self.init_scale))
 
-    def build_once(self, input, label, init_hidden, init_cell):
-        pass
-
     def forward(self, input, label, init_hidden, init_cell):
         init_h = fluid.layers.reshape(
             init_hidden, shape=[self.num_layers, -1, self.hidden_size])