MCPcopy
hub / github.com/tensorlayer/TensorLayer / build

Method build

tensorlayer/layers/embedding.py:230–265  ·  view source on GitHub ↗

Parameters ---------- inputs_shape : tuple the shape of inputs tensor

(self, inputs_shape)

Source from the content-addressed store, hash-verified

228 return s.format(classname=self.__class__.__name__, **self.__dict__)
229
230 def build(self, inputs_shape):
231 """
232 Parameters
233 ----------
234 inputs_shape : tuple
235 the shape of inputs tensor
236 """
237 # Look up embeddings for inputs.
238 # Note: a row of 'embeddings' is the vector representation of a word.
239 # for the sake of speed, it is better to slice the embedding matrix
240 # instead of transferring a word id to one-hot-format vector and then
241 # multiply by the embedding matrix.
242 # embed is the outputs of the hidden layer (embedding layer), it is a
243 # row vector with 'embedding_size' values.
244
245 self.embeddings = self._get_weights(
246 "embeddings",
247 shape=(self.vocabulary_size, self.embedding_size),
248 init=self.E_init,
249 )
250
251 self.normalized_embeddings = tf.nn.l2_normalize(self.embeddings, 1)
252
253 if self.activate_nce_loss:
254 # Construct the variables for the NCE loss (i.e. negative sampling)
255 self.nce_weights = self._get_weights(
256 "nce_weights",
257 shape=(self.vocabulary_size, self.embedding_size),
258 init=self.nce_W_init,
259 )
260
261 self.nce_biases = self._get_weights(
262 "nce_biases",
263 shape=(self.vocabulary_size, ),
264 init=self.nce_b_init,
265 )
266
267 # @tf.function
268 def forward(self, inputs, use_nce_loss=None):

Callers 1

__init__Method · 0.95

Calls 1

_get_weightsMethod · 0.80

Tested by

no test coverage detected