MCPcopy Create free account
hub / github.com/brightmart/text_classification / train

Method train

a08_EntityNetwork/a3_entity_network.py:242–250  ·  view source on GitHub ↗

based on the loss, use SGD to update parameter

(self)

Source from the content-addressed store, hash-verified

240 return xentropy - normalizing
241
242 def train(self):
243 """based on the loss, use SGD to update parameter"""
244 learning_rate = tf.train.exponential_decay(self.learning_rate, self.global_step, self.decay_steps,
245 self.decay_rate, staircase=True)
246 self.learning_rate_=learning_rate
247 #noise_std_dev = tf.constant(0.3) / (tf.sqrt(tf.cast(tf.constant(1) + self.global_step, tf.float32))) #gradient_noise_scale=noise_std_dev
248 train_op = tf_contrib.layers.optimize_loss(self.loss_val, global_step=self.global_step,
249 learning_rate=learning_rate, optimizer="Adam",clip_gradients=self.clip_gradients)
250 return train_op
251
252 #:param s_t: vector representation of current input(is a sentence). shape:[batch_size,sequence_length,embed_size]
253 #:param h: value(hidden state).shape:[hidden_size]

Callers 1

__init__Method · 0.95

Calls

no outgoing calls

Tested by

no test coverage detected