4ce9a9d078d41af1a960f0e5bf16e373f69d5117,open_seq2seq/encoders/rnn_encoders.py,UnidirectionalRNNEncoderWithEmbedding,_encode,#UnidirectionalRNNEncoderWithEmbedding#,71

Before Change

    cell_params = copy.deepcopy(self.params)
    cell_params["num_units"] = self.params["encoder_cell_units"]

    self._enc_emb_w = tf.get_variable(
      name="EncoderEmbeddingMatrix",
      shape=[self._src_vocab_size, self._src_emb_size],
      dtype=tf.float32
    )

    if self._mode == "train":
      dp_input_keep_prob = self.params["encoder_dp_input_keep_prob"]
      dp_output_keep_prob = self.params["encoder_dp_output_keep_prob"]
    else:
      dp_input_keep_prob = 1.0
      dp_output_keep_prob = 1.0

    self._encoder_cell_fw = create_rnn_cell(
      cell_type=self.params["encoder_cell_type"],
      cell_params=cell_params,
      num_layers=self.params["encoder_layers"],
      dp_input_keep_prob=dp_input_keep_prob,
      dp_output_keep_prob=dp_output_keep_prob,
      residual_connections=self.params["encoder_use_skip_connections"],
    )

    time_major = self.params.get("time_major", False)
    use_swap_memory = self.params.get("use_swap_memory", False)

After Change

      dp_input_keep_prob = 1.0
      dp_output_keep_prob = 1.0

    fwd_cells = [
      single_cell(cell_class=self.params["core_cell"],
                  cell_params=self.params.get("core_cell_params", {}),
                  dp_input_keep_prob=dp_input_keep_prob,
                  dp_output_keep_prob=dp_output_keep_prob,
                  residual_connections=self.params[
                    "encoder_use_skip_connections"]
                  ) for _ in range(self.params["encoder_layers"])]

    self._encoder_cell_fw = tf.contrib.rnn.MultiRNNCell(fwd_cells)

    time_major = self.params.get("time_major", False)
    use_swap_memory = self.params.get("use_swap_memory", False)

Instances