python/ops/helper.py

# Copyright 2016 The TensorFlow Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# ==============================================================================
"""A library of helpers for use with SamplingDecoders.
"""

from __future__ import absolute_import
from __future__ import division
from __future__ import print_function

import abc

import six

from tensorflow.contrib.seq2seq.python.ops import decoder
from tensorflow.python.framework import dtypes
from tensorflow.python.framework import ops
from tensorflow.python.framework import tensor_shape
from tensorflow.python.ops import array_ops
from tensorflow.python.ops import control_flow_ops
from tensorflow.python.ops import embedding_ops
from tensorflow.python.ops import gen_array_ops
from tensorflow.python.ops import math_ops
from tensorflow.python.ops import tensor_array_ops
from tensorflow.python.ops.distributions import bernoulli
from tensorflow.python.ops.distributions import categorical
from tensorflow.python.util import nest

__all__ = [
    "Helper",
    "TrainingHelper",
    "GreedyEmbeddingHelper",
    "SampleEmbeddingHelper",
    "CustomHelper",
    "ScheduledEmbeddingTrainingHelper",
    "ScheduledOutputTrainingHelper",
    "InferenceHelper",
]

_transpose_batch_time = decoder._transpose_batch_time  # pylint: disable=protected-access


def _unstack_ta(inp):
  return tensor_array_ops.TensorArray(
      dtype=inp.dtype, size=array_ops.shape(inp)[0],
      element_shape=inp.get_shape()[1:]).unstack(inp)


@six.add_metaclass(abc.ABCMeta)
class Helper(object):
  """Interface for implementing sampling in seq2seq decoders.

  Helper instances are used by `BasicDecoder`.
  """

  @abc.abstractproperty
  def batch_size(self):
    """Batch size of tensor returned by `sample`.

    Returns a scalar int32 tensor.
    """
    raise NotImplementedError("batch_size has not been implemented")

  @abc.abstractproperty
  def sample_ids_shape(self):
    """Shape of tensor returned by `sample`, excluding the batch dimension.

    Returns a `TensorShape`.
    """
    raise NotImplementedError("sample_ids_shape has not been implemented")

  @abc.abstractproperty
  def sample_ids_dtype(self):
    """DType of tensor returned by `sample`.

    Returns a DType.
    """
    raise NotImplementedError("sample_ids_dtype has not been implemented")

  @abc.abstractmethod
  def initialize(self, name=None):
    """Returns `(initial_finished, initial_inputs)`."""
    pass

  @abc.abstractmethod
  def sample(self, time, outputs, state, name=None):
    """Returns `sample_ids`."""
    pass

  @abc.abstractmethod
  def next_inputs(self, time, outputs, state, sample_ids, name=None):
    """Returns `(finished, next_inputs, next_state)`."""
    pass


class CustomHelper(Helper):
  """Base abstract class that allows the user to customize sampling."""

  def __init__(self, initialize_fn, sample_fn, next_inputs_fn,
               sample_ids_shape=None, sample_ids_dtype=None):
    """Initializer.

    Args:
      initialize_fn: callable that returns `(finished, next_inputs)`
        for the first iteration.
      sample_fn: callable that takes `(time, outputs, state)`
        and emits tensor `sample_ids`.
      next_inputs_fn: callable that takes `(time, outputs, state, sample_ids)`
        and emits `(finished, next_inputs, next_state)`.
      sample_ids_shape: Either a list of integers, or a 1-D Tensor of type
        `int32`, the shape of each value in the `sample_ids` batch. Defaults to
        a scalar.
      sample_ids_dtype: The dtype of the `sample_ids` tensor. Defaults to int32.
    """
    self._initialize_fn = initialize_fn
    self._sample_fn = sample_fn
    self._next_inputs_fn = next_inputs_fn
    self._batch_size = None
    self._sample_ids_shape = tensor_shape.TensorShape(sample_ids_shape or [])
    self._sample_ids_dtype = sample_ids_dtype or dtypes.int32

  @property
  def batch_size(self):
    if self._batch_size is None:
      raise ValueError("batch_size accessed before initialize was called")
    return self._batch_size

  @property
  def sample_ids_shape(self):
    return self._sample_ids_shape

  @property
  def sample_ids_dtype(self):
    return self._sample_ids_dtype

  def initialize(self, name=None):
    with ops.name_scope(name, "%sInitialize" % type(self).__name__):
      (finished, next_inputs) = self._initialize_fn()
      if self._batch_size is None:
        self._batch_size = array_ops.size(finished)
    return (finished, next_inputs)

  def sample(self, time, outputs, state, name=None):
    with ops.name_scope(
        name, "%sSample" % type(self).__name__, (time, outputs, state)):
      return self._sample_fn(time=time, outputs=outputs, state=state)

  def next_inputs(self, time, outputs, state, sample_ids, name=None):
    with ops.name_scope(
        name, "%sNextInputs" % type(self).__name__, (time, outputs, state)):
      return self._next_inputs_fn(
          time=time, outputs=outputs, state=state, sample_ids=sample_ids)


class TrainingHelper(Helper):
  """A helper for use during training.  Only reads inputs.

  Returned sample_ids are the argmax of the RNN output logits.
  """

  def __init__(self, inputs, sequence_length, time_major=False, name=None):
    """Initializer.

    Args:
      inputs: A (structure of) input tensors.
      sequence_length: An int32 vector tensor.
      time_major: Python bool.  Whether the tensors in `inputs` are time major.
        If `False` (default), they are assumed to be batch major.
      name: Name scope for any created operations.

    Raises:
      ValueError: if `sequence_length` is not a 1D tensor.
    """
    with ops.name_scope(name, "TrainingHelper", [inputs, sequence_length]):
      inputs = ops.convert_to_tensor(inputs, name="inputs")
      self._inputs = inputs
      if not time_major:
        inputs = nest.map_structure(_transpose_batch_time, inputs)

      self._input_tas = nest.map_structure(_unstack_ta, inputs)
      self._sequence_length = ops.convert_to_tensor(
          sequence_length, name="sequence_length")
      if self._sequence_length.get_shape().ndims != 1:
        raise ValueError(
            "Expected sequence_length to be a vector, but received shape: %s" %
            self._sequence_length.get_shape())

      self._zero_inputs = nest.map_structure(
          lambda inp: array_ops.zeros_like(inp[0, :]), inputs)

      self._batch_size = array_ops.size(sequence_length)

  @property
  def inputs(self):
    return self._inputs

  @property
  def sequence_length(self):
    return self._sequence_length

  @property
  def batch_size(self):
    return self._batch_size

  @property
  def sample_ids_shape(self):
    return tensor_shape.TensorShape([])

  @property
  def sample_ids_dtype(self):
    return dtypes.int32

  def initialize(self, name=None):
    with ops.name_scope(name, "TrainingHelperInitialize"):
      finished = math_ops.equal(0, self._sequence_length)
      all_finished = math_ops.reduce_all(finished)
      next_inputs = control_flow_ops.cond(
          all_finished, lambda: self._zero_inputs,
          lambda: nest.map_structure(lambda inp: inp.read(0), self._input_tas))
      return (finished, next_inputs)

  def sample(self, time, outputs, name=None, **unused_kwargs):
    with ops.name_scope(name, "TrainingHelperSample", [time, outputs]):
      sample_ids = math_ops.cast(
          math_ops.argmax(outputs, axis=-1), dtypes.int32)
      return sample_ids

  def next_inputs(self, time, outputs, state, name=None, **unused_kwargs):
    """next_inputs_fn for TrainingHelper."""
    with ops.name_scope(name, "TrainingHelperNextInputs",
                        [time, outputs, state]):
      next_time = time + 1
      finished = (next_time >= self._sequence_length)
      all_finished = math_ops.reduce_all(finished)
      def read_from_ta(inp):
        return inp.read(next_time)
      next_inputs = control_flow_ops.cond(
          all_finished, lambda: self._zero_inputs,
          lambda: nest.map_structure(read_from_ta, self._input_tas))
      return (finished, next_inputs, state)


class ScheduledEmbeddingTrainingHelper(TrainingHelper):
  """A training helper that adds scheduled sampling.

  Returns -1s for sample_ids where no sampling took place; valid sample id
  values elsewhere.
  """

  def __init__(self, inputs, sequence_length, embedding, sampling_probability,
               time_major=False, seed=None, scheduling_seed=None, name=None):
    """Initializer.

    Args:
      inputs: A (structure of) input tensors.
      sequence_length: An int32 vector tensor.
      embedding: A callable that takes a vector tensor of `ids` (argmax ids),
        or the `params` argument for `embedding_lookup`.
      sampling_probability: A 0D `float32` tensor: the probability of sampling
        categorically from the output ids instead of reading directly from the
        inputs.
      time_major: Python bool.  Whether the tensors in `inputs` are time major.
        If `False` (default), they are assumed to be batch major.
      seed: The sampling seed.
      scheduling_seed: The schedule decision rule sampling seed.
      name: Name scope for any created operations.

    Raises:
      ValueError: if `sampling_probability` is not a scalar or vector.
    """
    with ops.name_scope(name, "ScheduledEmbeddingSamplingWrapper",
                        [embedding, sampling_probability]):
      if callable(embedding):
        self._embedding_fn = embedding
      else:
        self._embedding_fn = (
            lambda ids: embedding_ops.embedding_lookup(embedding, ids))
      self._sampling_probability = ops.convert_to_tensor(
          sampling_probability, name="sampling_probability")
      if self._sampling_probability.get_shape().ndims not in (0, 1):
        raise ValueError(
            "sampling_probability must be either a scalar or a vector. "
            "saw shape: %s" % (self._sampling_probability.get_shape()))
      self._seed = seed
      self._scheduling_seed = scheduling_seed
      super(ScheduledEmbeddingTrainingHelper, self).__init__(
          inputs=inputs,
          sequence_length=sequence_length,
          time_major=time_major,
          name=name)

  def initialize(self, name=None):
    return super(ScheduledEmbeddingTrainingHelper, self).initialize(name=name)

  def sample(self, time, outputs, state, name=None):
    with ops.name_scope(name, "ScheduledEmbeddingTrainingHelperSample",
                        [time, outputs, state]):
      # Return -1s where we did not sample, and sample_ids elsewhere
      select_sampler = bernoulli.Bernoulli(
          probs=self._sampling_probability, dtype=dtypes.bool)
      select_sample = select_sampler.sample(
          sample_shape=self.batch_size, seed=self._scheduling_seed)
      sample_id_sampler = categorical.Categorical(logits=outputs)
      return array_ops.where(
          select_sample,
          sample_id_sampler.sample(seed=self._seed),
          gen_array_ops.fill([self.batch_size], -1))

  def next_inputs(self, time, outputs, state, sample_ids, name=None):
    with ops.name_scope(name, "ScheduledEmbeddingTrainingHelperNextInputs",
                        [time, outputs, state, sample_ids]):
      (finished, base_next_inputs, state) = (
          super(ScheduledEmbeddingTrainingHelper, self).next_inputs(
              time=time,
              outputs=outputs,
              state=state,
              sample_ids=sample_ids,
              name=name))

      def maybe_sample():
        """Perform scheduled sampling."""
        where_sampling = math_ops.cast(
            array_ops.where(sample_ids > -1), dtypes.int32)
        where_not_sampling = math_ops.cast(
            array_ops.where(sample_ids <= -1), dtypes.int32)
        sample_ids_sampling = array_ops.gather_nd(sample_ids, where_sampling)
        inputs_not_sampling = array_ops.gather_nd(
            base_next_inputs, where_not_sampling)
        sampled_next_inputs = self._embedding_fn(sample_ids_sampling)
        base_shape = array_ops.shape(base_next_inputs)
        return (array_ops.scatter_nd(indices=where_sampling,
                                     updates=sampled_next_inputs,
                                     shape=base_shape)
                + array_ops.scatter_nd(indices=where_not_sampling,
                                       updates=inputs_not_sampling,
                                       shape=base_shape))

      all_finished = math_ops.reduce_all(finished)
      next_inputs = control_flow_ops.cond(
          all_finished, lambda: base_next_inputs, maybe_sample)
      return (finished, next_inputs, state)


class ScheduledOutputTrainingHelper(TrainingHelper):
  """A training helper that adds scheduled sampling directly to outputs.

  Returns False for sample_ids where no sampling took place; True elsewhere.
  """

  def __init__(self, inputs, sequence_length, sampling_probability,
               time_major=False, seed=None, next_inputs_fn=None,
               auxiliary_inputs=None, name=None):
    """Initializer.

    Args:
      inputs: A (structure) of input tensors.
      sequence_length: An int32 vector tensor.
      sampling_probability: A 0D `float32` tensor: the probability of sampling
        from the outputs instead of reading directly from the inputs.
      time_major: Python bool.  Whether the tensors in `inputs` are time major.
        If `False` (default), they are assumed to be batch major.
      seed: The sampling seed.
      next_inputs_fn: (Optional) callable to apply to the RNN outputs to create
        the next input when sampling. If `None` (default), the RNN outputs will
        be used as the next inputs.
      auxiliary_inputs: An optional (structure of) auxiliary input tensors with
        a shape that matches `inputs` in all but (potentially) the final
        dimension. These tensors will be concatenated to the sampled output or
        the `inputs` when not sampling for use as the next input.
      name: Name scope for any created operations.

    Raises:
      ValueError: if `sampling_probability` is not a scalar or vector.
    """
    with ops.name_scope(name, "ScheduledOutputTrainingHelper",
                        [inputs, auxiliary_inputs, sampling_probability]):
      self._sampling_probability = ops.convert_to_tensor(
          sampling_probability, name="sampling_probability")
      if self._sampling_probability.get_shape().ndims not in (0, 1):
        raise ValueError(
            "sampling_probability must be either a scalar or a vector. "
            "saw shape: %s" % (self._sampling_probability.get_shape()))

      if auxiliary_inputs is None:
        maybe_concatenated_inputs = inputs
      else:
        inputs = ops.convert_to_tensor(inputs, name="inputs")
        auxiliary_inputs = ops.convert_to_tensor(
            auxiliary_inputs, name="auxiliary_inputs")
        maybe_concatenated_inputs = nest.map_structure(
            lambda x, y: array_ops.concat((x, y), -1),
            inputs, auxiliary_inputs)
        if not time_major:
          auxiliary_inputs = nest.map_structure(
              _transpose_batch_time, auxiliary_inputs)

      self._auxiliary_input_tas = (
          nest.map_structure(_unstack_ta, auxiliary_inputs)
          if auxiliary_inputs is not None else None)

      self._seed = seed

      self._next_inputs_fn = next_inputs_fn

      super(ScheduledOutputTrainingHelper, self).__init__(
          inputs=maybe_concatenated_inputs,
          sequence_length=sequence_length,
          time_major=time_major,
          name=name)

  def initialize(self, name=None):
    return super(ScheduledOutputTrainingHelper, self).initialize(name=name)

  def sample(self, time, outputs, state, name=None):
    with ops.name_scope(name, "ScheduledOutputTrainingHelperSample",
                        [time, outputs, state]):
      sampler = bernoulli.Bernoulli(probs=self._sampling_probability)
      return sampler.sample(sample_shape=self.batch_size, seed=self._seed)

  def next_inputs(self, time, outputs, state, sample_ids, name=None):
    with ops.name_scope(name, "ScheduledOutputTrainingHelperNextInputs",
                        [time, outputs, state, sample_ids]):
      (finished, base_next_inputs, state) = (
          super(ScheduledOutputTrainingHelper, self).next_inputs(
              time=time,
              outputs=outputs,
              state=state,
              sample_ids=sample_ids,
              name=name))
      sample_ids = math_ops.cast(sample_ids, dtypes.bool)

      def maybe_sample():
        """Perform scheduled sampling."""

        def maybe_concatenate_auxiliary_inputs(outputs_, indices=None):
          """Concatenate outputs with auxiliary inputs, if they exist."""
          if self._auxiliary_input_tas is None:
            return outputs_

          next_time = time + 1
          auxiliary_inputs = nest.map_structure(
              lambda ta: ta.read(next_time), self._auxiliary_input_tas)
          if indices is not None:
            auxiliary_inputs = array_ops.gather_nd(auxiliary_inputs, indices)
          return nest.map_structure(
              lambda x, y: array_ops.concat((x, y), -1),
              outputs_, auxiliary_inputs)

        if self._next_inputs_fn is None:
          return array_ops.where(
              sample_ids, maybe_concatenate_auxiliary_inputs(outputs),
              base_next_inputs)

        where_sampling = math_ops.cast(
            array_ops.where(sample_ids), dtypes.int32)
        where_not_sampling = math_ops.cast(
            array_ops.where(math_ops.logical_not(sample_ids)), dtypes.int32)
        outputs_sampling = array_ops.gather_nd(outputs, where_sampling)
        inputs_not_sampling = array_ops.gather_nd(base_next_inputs,
                                                  where_not_sampling)
        sampled_next_inputs = maybe_concatenate_auxiliary_inputs(
            self._next_inputs_fn(outputs_sampling), where_sampling)

        base_shape = array_ops.shape(base_next_inputs)
        return (array_ops.scatter_nd(indices=where_sampling,
                                     updates=sampled_next_inputs,
                                     shape=base_shape)
                + array_ops.scatter_nd(indices=where_not_sampling,
                                       updates=inputs_not_sampling,
                                       shape=base_shape))

      all_finished = math_ops.reduce_all(finished)
      no_samples = math_ops.logical_not(math_ops.reduce_any(sample_ids))
      next_inputs = control_flow_ops.cond(
          math_ops.logical_or(all_finished, no_samples),
          lambda: base_next_inputs, maybe_sample)
      return (finished, next_inputs, state)


class GreedyEmbeddingHelper(Helper):
  """A helper for use during inference.

  Uses the argmax of the output (treated as logits) and passes the
  result through an embedding layer to get the next input.
  """

  def __init__(self, embedding, start_tokens, end_token):
    """Initializer.

    Args:
      embedding: A callable that takes a vector tensor of `ids` (argmax ids),
        or the `params` argument for `embedding_lookup`. The returned tensor
        will be passed to the decoder input.
      start_tokens: `int32` vector shaped `[batch_size]`, the start tokens.
      end_token: `int32` scalar, the token that marks end of decoding.

    Raises:
      ValueError: if `start_tokens` is not a 1D tensor or `end_token` is not a
        scalar.
    """
    if callable(embedding):
      self._embedding_fn = embedding
    else:
      self._embedding_fn = (
          lambda ids: embedding_ops.embedding_lookup(embedding, ids))

    self._start_tokens = ops.convert_to_tensor(
        start_tokens, dtype=dtypes.int32, name="start_tokens")
    self._end_token = ops.convert_to_tensor(
        end_token, dtype=dtypes.int32, name="end_token")
    if self._start_tokens.get_shape().ndims != 1:
      raise ValueError("start_tokens must be a vector")
    self._batch_size = array_ops.size(start_tokens)
    if self._end_token.get_shape().ndims != 0:
      raise ValueError("end_token must be a scalar")
    self._start_inputs = self._embedding_fn(self._start_tokens)

  @property
  def batch_size(self):
    return self._batch_size

  @property
  def sample_ids_shape(self):
    return tensor_shape.TensorShape([])

  @property
  def sample_ids_dtype(self):
    return dtypes.int32

  def initialize(self, name=None):
    finished = array_ops.tile([False], [self._batch_size])
    return (finished, self._start_inputs)

  def sample(self, time, outputs, state, name=None):
    """sample for GreedyEmbeddingHelper."""
    del time, state  # unused by sample_fn
    # Outputs are logits, use argmax to get the most probable id
    if not isinstance(outputs, ops.Tensor):
      raise TypeError("Expected outputs to be a single Tensor, got: %s" %
                      type(outputs))
    sample_ids = math_ops.argmax(outputs, axis=-1, output_type=dtypes.int32)
    return sample_ids

  def next_inputs(self, time, outputs, state, sample_ids, name=None):
    """next_inputs_fn for GreedyEmbeddingHelper."""
    del time, outputs  # unused by next_inputs_fn
    finished = math_ops.equal(sample_ids, self._end_token)
    all_finished = math_ops.reduce_all(finished)
    next_inputs = control_flow_ops.cond(
        all_finished,
        # If we're finished, the next_inputs value doesn't matter
        lambda: self._start_inputs,
        lambda: self._embedding_fn(sample_ids))
    return (finished, next_inputs, state)


class SampleEmbeddingHelper(GreedyEmbeddingHelper):
  """A helper for use during inference.

  Uses sampling (from a distribution) instead of argmax and passes the
  result through an embedding layer to get the next input.
  """

  def __init__(self, embedding, start_tokens, end_token,
               softmax_temperature=None, seed=None):
    """Initializer.

    Args:
      embedding: A callable that takes a vector tensor of `ids` (argmax ids),
        or the `params` argument for `embedding_lookup`. The returned tensor
        will be passed to the decoder input.
      start_tokens: `int32` vector shaped `[batch_size]`, the start tokens.
      end_token: `int32` scalar, the token that marks end of decoding.
      softmax_temperature: (Optional) `float32` scalar, value to divide the
        logits by before computing the softmax. Larger values (above 1.0) result
        in more random samples, while smaller values push the sampling
        distribution towards the argmax. Must be strictly greater than 0.
        Defaults to 1.0.
      seed: (Optional) The sampling seed.

    Raises:
      ValueError: if `start_tokens` is not a 1D tensor or `end_token` is not a
        scalar.
    """
    super(SampleEmbeddingHelper, self).__init__(
        embedding, start_tokens, end_token)
    self._softmax_temperature = softmax_temperature
    self._seed = seed

  def sample(self, time, outputs, state, name=None):
    """sample for SampleEmbeddingHelper."""
    del time, state  # unused by sample_fn
    # Outputs are logits, we sample instead of argmax (greedy).
    if not isinstance(outputs, ops.Tensor):
      raise TypeError("Expected outputs to be a single Tensor, got: %s" %
                      type(outputs))
    if self._softmax_temperature is None:
      logits = outputs
    else:
      logits = outputs / self._softmax_temperature

    sample_id_sampler = categorical.Categorical(logits=logits)
    sample_ids = sample_id_sampler.sample(seed=self._seed)

    return sample_ids


class InferenceHelper(Helper):
  """A helper to use during inference with a custom sampling function."""

  def __init__(self, sample_fn, sample_shape, sample_dtype,
               start_inputs, end_fn, next_inputs_fn=None):
    """Initializer.

    Args:
      sample_fn: A callable that takes `outputs` and emits tensor `sample_ids`.
      sample_shape: Either a list of integers, or a 1-D Tensor of type `int32`,
        the shape of the each sample in the batch returned by `sample_fn`.
      sample_dtype: the dtype of the sample returned by `sample_fn`.
      start_inputs: The initial batch of inputs.
      end_fn: A callable that takes `sample_ids` and emits a `bool` vector
        shaped `[batch_size]` indicating whether each sample is an end token.
      next_inputs_fn: (Optional) A callable that takes `sample_ids` and returns
        the next batch of inputs. If not provided, `sample_ids` is used as the
        next batch of inputs.
    """
    self._sample_fn = sample_fn
    self._end_fn = end_fn
    self._sample_shape = tensor_shape.TensorShape(sample_shape)
    self._sample_dtype = sample_dtype
    self._next_inputs_fn = next_inputs_fn
    self._batch_size = array_ops.shape(start_inputs)[0]
    self._start_inputs = ops.convert_to_tensor(
        start_inputs, name="start_inputs")

  @property
  def batch_size(self):
    return self._batch_size

  @property
  def sample_ids_shape(self):
    return self._sample_shape

  @property
  def sample_ids_dtype(self):
    return self._sample_dtype

  def initialize(self, name=None):
    finished = array_ops.tile([False], [self._batch_size])
    return (finished, self._start_inputs)

  def sample(self, time, outputs, state, name=None):
    del time, state  # unused by sample
    return self._sample_fn(outputs)

  def next_inputs(self, time, outputs, state, sample_ids, name=None):
    del time, outputs  # unused by next_inputs
    if self._next_inputs_fn is None:
      next_inputs = sample_ids
    else:
      next_inputs = self._next_inputs_fn(sample_ids)
    finished = self._end_fn(sample_ids)
    return (finished, next_inputs, state)