rnn_encoder.py 文件源码

def encode(self, inputs, sequence_length, **kwargs):
    scope = tf.get_variable_scope()
    scope.set_initializer(tf.random_uniform_initializer(
        -self.params["init_scale"],
        self.params["init_scale"]))

    cell_fw = training_utils.get_rnn_cell(**self.params["rnn_cell"])
    cell_bw = training_utils.get_rnn_cell(**self.params["rnn_cell"])
    outputs, states = tf.nn.bidirectional_dynamic_rnn(
        cell_fw=cell_fw,
        cell_bw=cell_bw,
        inputs=inputs,
        sequence_length=sequence_length,
        dtype=tf.float32,
        **kwargs)

    # Concatenate outputs and states of the forward and backward RNNs
    outputs_concat = tf.concat(outputs, 2)

    return EncoderOutput(
        outputs=outputs_concat,
        final_state=states,
        attention_values=outputs_concat,
        attention_values_length=sequence_length)