Function _reverse_seq

DIEN/rnn.py:268–307

Reverse a list of Tensors up to specified lengths. Args: input_seq: Sequence of seq_len tensors of dimension (batch_size, n_features) or nested tuples of tensors. lengths: A `Tensor` of dimension batch_size, containing lengths for each sequence in the batch

(input_seq, lengths)

def _reverse_seq(input_seq, lengths):
  """Reverse a list of Tensors up to specified lengths.

  Args:
    input_seq: Sequence of seq_len tensors of dimension (batch_size, n_features)
               or nested tuples of tensors.
    lengths:   A `Tensor` of dimension batch_size, containing lengths for each
               sequence in the batch. If "None" is specified, simply reverses
               the list.

  Returns:
    time-reversed sequence
  """
  if lengths is None:
    return list(reversed(input_seq))

  flat_input_seq = tuple(nest.flatten(input_) for input_ in input_seq)

  flat_results = [[] for _ in range(len(input_seq))]
  for sequence in zip(*flat_input_seq):
    input_shape = tensor_shape.unknown_shape(
        ndims=sequence[0].get_shape().ndims)
    for input_ in sequence:
      input_shape.merge_with(input_.get_shape())
      input_.set_shape(input_shape)

    # Join into (time, batch_size, depth)
    s_joined = array_ops.stack(sequence)

    # Reverse along dimension 0
    s_reversed = array_ops.reverse_sequence(s_joined, lengths, 0, 1)
    # Split again into list
    result = array_ops.unstack(s_reversed)
    for r, flat_result in zip(result, flat_results):
      r.set_shape(input_shape)
      flat_result.append(r)

  results = [nest.pack_sequence_as(structure=input_, flat_sequence=flat_result)
             for input_, flat_result in zip(input_seq, flat_results)]
  return results


def bidirectional_dynamic_rnn(cell_fw, cell_bw, inputs, sequence_length=None,