INNER CODE UNIT · Python

learning_rate

ematvey/hierarchical-attention-networks · HAN_model.py:24

               learning_rate=1e-4,
               device='/cpu:0',
               scope=None):
    self.vocab_size = vocab_size
    self.embedding_size = embedding_size
    self.classes = classes
    self.word_cell = word_cell
    self.word_output_size = word_output_size
    self.sentence_cell = sentence_cell
    self.sentence_output_size = sentence_output_size
    self.max_grad_norm = max_grad_norm
    self.dropout_keep_proba = dropout_keep_proba

    with tf.variable_scope(scope or 'tcm') as scope:
      self.global_step = tf.Variable(0, name='global_step', trainable=False)

      if is_training is not None:
        self.is_training = is_training

View source record →