3497526703e85981f39b643e923dcb1e40eec366,texar/modules/encoders/rnn_encoders.py,BidirectionalRNNEncoder,_build,#BidirectionalRNNEncoder#,712

Before Change

                time_major=time_major,
                **kwargs)

        map_func_fw = functools.partial(
            _forward_output_layers,
            output_layer=self._output_layer_fw,
            time_major=time_major,
            hparams=self._hparams.output_layer_fw,
            mode=mode)
        outputs_fw = nest.map_structure(map_func_fw, cell_outputs[0])

        hparams_output_layer_bw = self._hparams.output_layer_bw
        if self._hparams.output_layer_share_config:
            hparams_output_layer_bw = self._hparams.output_layer_fw
        map_func_bw = functools.partial(
            _forward_output_layers,
            output_layer=self._output_layer_bw,
            time_major=time_major,
            hparams=hparams_output_layer_bw,
            mode=mode)
        outputs_bw = nest.map_structure(map_func_bw, cell_outputs[1])

        outputs = (outputs_fw, outputs_bw)

        if not self._built:
            self._add_internal_trainable_variables()
            // Add trainable variables of cells and output layers
            // which may be constructed externally.
            self._add_trainable_variable(
                layers.get_rnn_cell_trainable_variables(self._cell_fw))
            self._add_trainable_variable(
                layers.get_rnn_cell_trainable_variables(self._cell_bw))
            if self._output_layer_fw and \
                    not isinstance(self._output_layer_fw, (list, tuple)):
                self._add_trainable_variable(
                    self._output_layer_fw.trainable_variables)
            if self._output_layer_bw and \
                    not isinstance(self._output_layer_bw, (list, tuple)):
                self._add_trainable_variable(
                    self._output_layer_bw.trainable_variables)
            self._built = True

        if return_cell_output:
            return outputs, output_states, cell_outputs
        else:
            return outputs, output_states

    @staticmethod
    def concat_outputs(outputs):
        Concats the outputs of the bidirectional encoder into a single
        tensor.

After Change

                time_major=time_major,
                **kwargs)

        outputs_fw, output_size_fw = _apply_rnn_encoder_output_layer(
            self._output_layer_fw, time_major, self._output_layer_hparams_fw,
            mode, cell_outputs[0], self._cell_fw.output_size)

        outputs_bw, output_size_bw = _apply_rnn_encoder_output_layer(
            self._output_layer_bw, time_major, self._output_layer_hparams_bw,
            mode, cell_outputs[1], self._cell_bw.output_size)

Instances