Python tensorflow.python.ops.data_flow_ops.RandomShuffleQueue() Examples

def shuffle_join(tensor_list_list, capacity,
                 min_ad, phase):
    name = 'shuffel_input'
    types = _dtypes(tensor_list_list)
    queue = data_flow_ops.RandomShuffleQueue(
        capacity=capacity, min_after_dequeue=min_ad,
        dtypes=types)

    # Build enque Operations
    _enqueue_join(queue, tensor_list_list)

    full = (math_ops.cast(math_ops.maximum(0, queue.size() - min_ad),
                          dtypes.float32) * (1. / (capacity - min_ad)))
    # Note that name contains a '/' at the end so we intentionally do not place
    # a '/' after %s below.
    summary_name = (
        "queue/%s/fraction_over_%d_of_%d_full" %
        (name + '_' + phase, min_ad, capacity - min_ad))
    tf.summary.scalar(summary_name, full)

    dequeued = queue.dequeue(name='shuffel_deqeue')
    # dequeued = _deserialize_sparse_tensors(dequeued, sparse_info)
    return dequeued 

def shuffle_join(tensor_list_list, capacity,
                 min_ad, phase):
    name = 'shuffel_input'
    types = _dtypes(tensor_list_list)
    queue = data_flow_ops.RandomShuffleQueue(
        capacity=capacity, min_after_dequeue=min_ad,
        dtypes=types)

    # Build enque Operations
    _enqueue_join(queue, tensor_list_list)

    full = (math_ops.cast(math_ops.maximum(0, queue.size() - min_ad),
                          dtypes.float32) * (1. / (capacity - min_ad)))
    # Note that name contains a '/' at the end so we intentionally do not place
    # a '/' after %s below.
    summary_name = (
        "queue/%s/fraction_over_%d_of_%d_full" %
        (name + '_' + phase, min_ad, capacity - min_ad))
    tf.summary.scalar(summary_name, full)

    dequeued = queue.dequeue(name='shuffel_deqeue')
    # dequeued = _deserialize_sparse_tensors(dequeued, sparse_info)
    return dequeued 

def shuffle_join(tensor_list_list, capacity,
                 min_ad, phase):
    name = 'shuffel_input'
    types = _dtypes(tensor_list_list)
    queue = data_flow_ops.RandomShuffleQueue(
        capacity=capacity, min_after_dequeue=min_ad,
        dtypes=types)

    # Build enque Operations
    _enqueue_join(queue, tensor_list_list)

    full = (math_ops.cast(math_ops.maximum(0, queue.size() - min_ad),
                          dtypes.float32) * (1. / (capacity - min_ad)))
    # Note that name contains a '/' at the end so we intentionally do not place
    # a '/' after %s below.
    summary_name = (
        "queue/%s/fraction_over_%d_of_%d_full" %
        (name + '_' + phase, min_ad, capacity - min_ad))
    tf.summary.scalar(summary_name, full)

    dequeued = queue.dequeue(name='shuffel_deqeue')
    # dequeued = _deserialize_sparse_tensors(dequeued, sparse_info)
    return dequeued 

def _shuffle_batch(tensors, batch_size, capacity, min_after_dequeue,
                   keep_input, num_threads=1, seed=None, enqueue_many=False,
                   shapes=None, allow_smaller_final_batch=False,
                   shared_name=None, name=None):
  """Helper function for `shuffle_batch` and `maybe_shuffle_batch`."""
  tensor_list = _as_tensor_list(tensors)
  with ops.name_scope(name, "shuffle_batch",
                      list(tensor_list) + [keep_input]) as name:
    tensor_list = _validate(tensor_list)
    keep_input = _validate_keep_input(keep_input, enqueue_many)
    tensor_list, sparse_info = _store_sparse_tensors(
        tensor_list, enqueue_many, keep_input)
    types = _dtypes([tensor_list])
    shapes = _shapes([tensor_list], shapes, enqueue_many)
    queue = data_flow_ops.RandomShuffleQueue(
        capacity=capacity, min_after_dequeue=min_after_dequeue, seed=seed,
        dtypes=types, shapes=shapes, shared_name=shared_name)
    _enqueue(queue, tensor_list, num_threads, enqueue_many, keep_input)
    full = (math_ops.cast(math_ops.maximum(0, queue.size() - min_after_dequeue),
                          dtypes.float32) *
            (1. / (capacity - min_after_dequeue)))
    # Note that name contains a '/' at the end so we intentionally do not place
    # a '/' after %s below.
    summary_name = (
        "fraction_over_%d_of_%d_full" %
        (min_after_dequeue, capacity - min_after_dequeue))
    summary.scalar(summary_name, full)

    if allow_smaller_final_batch:
      dequeued = queue.dequeue_up_to(batch_size, name=name)
    else:
      dequeued = queue.dequeue_many(batch_size, name=name)
    dequeued = _restore_sparse_tensors(dequeued, sparse_info)
    return _as_original_type(tensors, dequeued) 

def _shuffle_batch_join(tensors_list, batch_size, capacity,
                        min_after_dequeue, keep_input, seed=None,
                        enqueue_many=False, shapes=None,
                        allow_smaller_final_batch=False, shared_name=None,
                        name=None):
  """Helper function for `shuffle_batch_join` and `maybe_shuffle_batch_join`."""
  tensor_list_list = _as_tensor_list_list(tensors_list)
  with ops.name_scope(name, "shuffle_batch_join",
                      _flatten(tensor_list_list) + [keep_input]) as name:
    tensor_list_list = _validate_join(tensor_list_list)
    keep_input = _validate_keep_input(keep_input, enqueue_many)
    tensor_list_list, sparse_info = _store_sparse_tensors_join(
        tensor_list_list, enqueue_many, keep_input)
    types = _dtypes(tensor_list_list)
    shapes = _shapes(tensor_list_list, shapes, enqueue_many)
    queue = data_flow_ops.RandomShuffleQueue(
        capacity=capacity, min_after_dequeue=min_after_dequeue, seed=seed,
        dtypes=types, shapes=shapes, shared_name=shared_name)
    _enqueue_join(queue, tensor_list_list, enqueue_many, keep_input)
    full = (math_ops.cast(math_ops.maximum(0, queue.size() - min_after_dequeue),
                          dtypes.float32) *
            (1. / (capacity - min_after_dequeue)))
    # Note that name contains a '/' at the end so we intentionally do not place
    # a '/' after %s below.
    summary_name = (
        "fraction_over_%d_of_%d_full" %
        (min_after_dequeue, capacity - min_after_dequeue))
    summary.scalar(summary_name, full)

    if allow_smaller_final_batch:
      dequeued = queue.dequeue_up_to(batch_size, name=name)
    else:
      dequeued = queue.dequeue_many(batch_size, name=name)
    dequeued = _restore_sparse_tensors(dequeued, sparse_info)
    # tensors_list was validated to not be empty.
    return _as_original_type(tensors_list[0], dequeued)

# Batching functions ---------------------------------------------------------- 

def _shuffle_batch(tensors, batch_size, capacity, min_after_dequeue,
                   keep_input, num_threads=1, seed=None, enqueue_many=False,
                   shapes=None, allow_smaller_final_batch=False,
                   shared_name=None, name=None):
  """Helper function for `shuffle_batch` and `maybe_shuffle_batch`."""
  tensor_list = _as_tensor_list(tensors)
  with ops.name_scope(name, "shuffle_batch",
                      list(tensor_list) + [keep_input]) as name:
    tensor_list = _validate(tensor_list)
    keep_input = _validate_tensor_or_none(keep_input)
    tensor_list, sparse_info = _store_sparse_tensors(
        tensor_list, enqueue_many, keep_input)
    types = _dtypes([tensor_list])
    shapes = _shapes([tensor_list], shapes, enqueue_many)
    queue = data_flow_ops.RandomShuffleQueue(
        capacity=capacity, min_after_dequeue=min_after_dequeue, seed=seed,
        dtypes=types, shapes=shapes, shared_name=shared_name)
    _enqueue(queue, tensor_list, num_threads, enqueue_many, keep_input)
    full = (math_ops.cast(math_ops.maximum(0, queue.size() - min_after_dequeue),
                          dtypes.float32) *
            (1. / (capacity - min_after_dequeue)))
    # Note that name contains a '/' at the end so we intentionally do not place
    # a '/' after %s below.
    summary_name = (
        "fraction_over_%d_of_%d_full" %
        (min_after_dequeue, capacity - min_after_dequeue))
    summary.scalar(summary_name, full)

    if allow_smaller_final_batch:
      dequeued = queue.dequeue_up_to(batch_size, name=name)
    else:
      dequeued = queue.dequeue_many(batch_size, name=name)
    dequeued = _restore_sparse_tensors(dequeued, sparse_info)
    return _as_original_type(tensors, dequeued) 

def _shuffle_batch_join(tensors_list, batch_size, capacity,
                        min_after_dequeue, keep_input, seed=None,
                        enqueue_many=False, shapes=None,
                        allow_smaller_final_batch=False, shared_name=None,
                        name=None):
  """Helper function for `shuffle_batch_join` and `maybe_shuffle_batch_join`."""
  tensor_list_list = _as_tensor_list_list(tensors_list)
  with ops.name_scope(name, "shuffle_batch_join",
                      _flatten(tensor_list_list) + [keep_input]) as name:
    tensor_list_list = _validate_join(tensor_list_list)
    keep_input = _validate_tensor_or_none(keep_input)
    tensor_list_list, sparse_info = _store_sparse_tensors_join(
        tensor_list_list, enqueue_many, keep_input)
    types = _dtypes(tensor_list_list)
    shapes = _shapes(tensor_list_list, shapes, enqueue_many)
    queue = data_flow_ops.RandomShuffleQueue(
        capacity=capacity, min_after_dequeue=min_after_dequeue, seed=seed,
        dtypes=types, shapes=shapes, shared_name=shared_name)
    _enqueue_join(queue, tensor_list_list, enqueue_many, keep_input)
    full = (math_ops.cast(math_ops.maximum(0, queue.size() - min_after_dequeue),
                          dtypes.float32) *
            (1. / (capacity - min_after_dequeue)))
    # Note that name contains a '/' at the end so we intentionally do not place
    # a '/' after %s below.
    summary_name = (
        "fraction_over_%d_of_%d_full" %
        (min_after_dequeue, capacity - min_after_dequeue))
    summary.scalar(summary_name, full)

    if allow_smaller_final_batch:
      dequeued = queue.dequeue_up_to(batch_size, name=name)
    else:
      dequeued = queue.dequeue_many(batch_size, name=name)
    dequeued = _restore_sparse_tensors(dequeued, sparse_info)
    # tensors_list was validated to not be empty.
    return _as_original_type(tensors_list[0], dequeued)

# Batching functions ---------------------------------------------------------- 

def testRandomShuffleQueue(self):
    shared_queue = data_flow_ops.RandomShuffleQueue(
        capacity=256,
        min_after_dequeue=128,
        dtypes=[dtypes_lib.string, dtypes_lib.string])
    self._verify_all_data_sources_read(shared_queue) 

def testRandomShuffleQueue(self):
    shared_queue = data_flow_ops.RandomShuffleQueue(
        capacity=256,
        min_after_dequeue=128,
        dtypes=[dtypes_lib.string, dtypes_lib.string])
    self._verify_all_data_sources_read(shared_queue) 

def testRandomShuffleQueue(self):
    shared_queue = data_flow_ops.RandomShuffleQueue(
        capacity=256,
        min_after_dequeue=128,
        dtypes=[dtypes_lib.string, dtypes_lib.string])
    self._verify_all_data_sources_read(shared_queue) 

def testReadUpToFromRandomShuffleQueue(self):
    shared_queue = data_flow_ops.RandomShuffleQueue(
        capacity=55,
        min_after_dequeue=28,
        dtypes=[dtypes_lib.string, dtypes_lib.string],
        shapes=[[], []])
    self._verify_read_up_to_out(shared_queue) 

def _shuffle_batch_join(tensors_list, batch_size, capacity,
                        min_after_dequeue, keep_input, seed=None,
                        enqueue_many=False, shapes=None,
                        allow_smaller_final_batch=False, shared_name=None,
                        name=None):
  """Helper function for `shuffle_batch_join` and `maybe_shuffle_batch_join`."""
  tensor_list_list = _as_tensor_list_list(tensors_list)
  with ops.name_scope(name, "shuffle_batch_join",
                      _flatten(tensor_list_list) + [keep_input]) as name:
    tensor_list_list = _validate_join(tensor_list_list)
    keep_input = _validate_tensor_or_none(keep_input)
    tensor_list_list, sparse_info = _store_sparse_tensors_join(
        tensor_list_list, enqueue_many, keep_input)
    types = _dtypes(tensor_list_list)
    shapes = _shapes(tensor_list_list, shapes, enqueue_many)
    queue = data_flow_ops.RandomShuffleQueue(
        capacity=capacity, min_after_dequeue=min_after_dequeue, seed=seed,
        dtypes=types, shapes=shapes, shared_name=shared_name)
    _enqueue_join(queue, tensor_list_list, enqueue_many, keep_input)
    full = (math_ops.cast(math_ops.maximum(0, queue.size() - min_after_dequeue),
                          dtypes.float32) *
            (1. / (capacity - min_after_dequeue)))
    # Note that name contains a '/' at the end so we intentionally do not place
    # a '/' after %s below.
    summary_name = (
        "fraction_over_%d_of_%d_full" %
        (min_after_dequeue, capacity - min_after_dequeue))
    summary.scalar(summary_name, full)

    if allow_smaller_final_batch:
      dequeued = queue.dequeue_up_to(batch_size, name=name)
    else:
      dequeued = queue.dequeue_many(batch_size, name=name)
    dequeued = _restore_sparse_tensors(dequeued, sparse_info)
    # tensors_list was validated to not be empty.
    return _as_original_type(tensors_list[0], dequeued)

# Batching functions ---------------------------------------------------------- 

def _shuffle_batch(tensors, batch_size, capacity, min_after_dequeue,
                   keep_input, num_threads=1, seed=None, enqueue_many=False,
                   shapes=None, allow_smaller_final_batch=False,
                   shared_name=None, name=None):
  """Helper function for `shuffle_batch` and `maybe_shuffle_batch`."""
  tensor_list = _as_tensor_list(tensors)
  with ops.name_scope(name, "shuffle_batch",
                      list(tensor_list) + [keep_input]) as name:
    tensor_list = _validate(tensor_list)
    keep_input = _validate_tensor_or_none(keep_input)
    tensor_list, sparse_info = _store_sparse_tensors(
        tensor_list, enqueue_many, keep_input)
    types = _dtypes([tensor_list])
    shapes = _shapes([tensor_list], shapes, enqueue_many)
    queue = data_flow_ops.RandomShuffleQueue(
        capacity=capacity, min_after_dequeue=min_after_dequeue, seed=seed,
        dtypes=types, shapes=shapes, shared_name=shared_name)
    _enqueue(queue, tensor_list, num_threads, enqueue_many, keep_input)
    full = (math_ops.cast(math_ops.maximum(0, queue.size() - min_after_dequeue),
                          dtypes.float32) *
            (1. / (capacity - min_after_dequeue)))
    # Note that name contains a '/' at the end so we intentionally do not place
    # a '/' after %s below.
    summary_name = (
        "fraction_over_%d_of_%d_full" %
        (min_after_dequeue, capacity - min_after_dequeue))
    summary.scalar(summary_name, full)

    if allow_smaller_final_batch:
      dequeued = queue.dequeue_up_to(batch_size, name=name)
    else:
      dequeued = queue.dequeue_many(batch_size, name=name)
    dequeued = _restore_sparse_tensors(dequeued, sparse_info)
    return _as_original_type(tensors, dequeued) 

def _shuffle_batch(tensors, batch_size, capacity, min_after_dequeue,
                   keep_input, num_threads=1, seed=None, enqueue_many=False,
                   shapes=None, allow_smaller_final_batch=False,
                   shared_name=None, name=None):
  """Helper function for `shuffle_batch` and `maybe_shuffle_batch`."""
  tensor_list = _as_tensor_list(tensors)
  with ops.name_scope(name, "shuffle_batch",
                      list(tensor_list) + [keep_input]) as name:
    if capacity <= min_after_dequeue:
      raise ValueError("capacity %d must be bigger than min_after_dequeue %d."
                       % (capacity, min_after_dequeue))
    tensor_list = _validate(tensor_list)
    keep_input = _validate_keep_input(keep_input, enqueue_many)
    tensor_list, sparse_info = _store_sparse_tensors(
        tensor_list, enqueue_many, keep_input)
    types = _dtypes([tensor_list])
    shapes = _shapes([tensor_list], shapes, enqueue_many)
    queue = data_flow_ops.RandomShuffleQueue(
        capacity=capacity, min_after_dequeue=min_after_dequeue, seed=seed,
        dtypes=types, shapes=shapes, shared_name=shared_name)
    _enqueue(queue, tensor_list, num_threads, enqueue_many, keep_input)
    full = (math_ops.to_float(
        math_ops.maximum(0, queue.size() - min_after_dequeue)) *
            (1. / (capacity - min_after_dequeue)))
    # Note that name contains a '/' at the end so we intentionally do not place
    # a '/' after %s below.
    summary_name = (
        "fraction_over_%d_of_%d_full" %
        (min_after_dequeue, capacity - min_after_dequeue))
    summary.scalar(summary_name, full)

    if allow_smaller_final_batch:
      dequeued = queue.dequeue_up_to(batch_size, name=name)
    else:
      dequeued = queue.dequeue_many(batch_size, name=name)
    dequeued = _restore_sparse_tensors(dequeued, sparse_info)
    return _as_original_type(tensors, dequeued) 

def _shuffle_batch_join(tensors_list, batch_size, capacity,
                        min_after_dequeue, keep_input, seed=None,
                        enqueue_many=False, shapes=None,
                        allow_smaller_final_batch=False, shared_name=None,
                        name=None):
  """Helper function for `shuffle_batch_join` and `maybe_shuffle_batch_join`."""
  tensor_list_list = _as_tensor_list_list(tensors_list)
  with ops.name_scope(name, "shuffle_batch_join",
                      _flatten(tensor_list_list) + [keep_input]) as name:
    tensor_list_list = _validate_join(tensor_list_list)
    keep_input = _validate_keep_input(keep_input, enqueue_many)
    tensor_list_list, sparse_info = _store_sparse_tensors_join(
        tensor_list_list, enqueue_many, keep_input)
    types = _dtypes(tensor_list_list)
    shapes = _shapes(tensor_list_list, shapes, enqueue_many)
    queue = data_flow_ops.RandomShuffleQueue(
        capacity=capacity, min_after_dequeue=min_after_dequeue, seed=seed,
        dtypes=types, shapes=shapes, shared_name=shared_name)
    _enqueue_join(queue, tensor_list_list, enqueue_many, keep_input)
    full = (math_ops.to_float(
        math_ops.maximum(0, queue.size() - min_after_dequeue)) *
            (1. / (capacity - min_after_dequeue)))
    # Note that name contains a '/' at the end so we intentionally do not place
    # a '/' after %s below.
    summary_name = (
        "fraction_over_%d_of_%d_full" %
        (min_after_dequeue, capacity - min_after_dequeue))
    summary.scalar(summary_name, full)

    if allow_smaller_final_batch:
      dequeued = queue.dequeue_up_to(batch_size, name=name)
    else:
      dequeued = queue.dequeue_many(batch_size, name=name)
    dequeued = _restore_sparse_tensors(dequeued, sparse_info)
    # tensors_list was validated to not be empty.
    return _as_original_type(tensors_list[0], dequeued)

# Batching functions ---------------------------------------------------------- 

def __init__(self,
               reader_class,
               common_queue,
               num_readers=4,
               reader_kwargs=None):
    """ParallelReader creates num_readers instances of the reader_class.

    Each instance is created by calling the `reader_class` function passing
    the arguments specified in `reader_kwargs` as in:
      reader_class(**read_kwargs)

    When you read from a ParallelReader, with its `read()` method,
    you just dequeue examples from the `common_queue`.

    The readers will read different files in parallel, asynchronously enqueueing
    their output into `common_queue`. The `common_queue.dtypes` must be
    [tf.string, tf.string]

    Because each reader can read from a different file, the examples in the
    `common_queue` could be from different files. Due to the asynchronous
    reading there is no guarantee that all the readers will read the same
    number of examples.

    If the `common_queue` is a shuffling queue, then the examples are shuffled.

    Usage:
      common_queue = tf.RandomShuffleQueue(
          capacity=256,
          min_after_dequeue=128,
          dtypes=[tf.string, tf.string])
      p_reader = ParallelReader(tf.TFRecordReader, common_queue)

      common_queue = tf.FIFOQueue(
          capacity=256,
          dtypes=[tf.string, tf.string])
      p_reader = ParallelReader(readers, common_queue, num_readers=2)


    Args:
      reader_class: one of the io_ops.ReaderBase subclasses ex: TFRecordReader
      common_queue: a Queue to hold (key, value pairs) with `dtypes` equal to
        [tf.string, tf.string]. Must be one of the data_flow_ops.Queues
        instances, ex. `tf.FIFOQueue()`, `tf.RandomShuffleQueue()`, ...
      num_readers: a integer, number of instances of reader_class to create.
      reader_kwargs: an optional dict of kwargs to create the readers.

    Raises:
      TypeError: if `common_queue.dtypes` is not [tf.string, tf.string].
    """
    if len(common_queue.dtypes) != 2:
      raise TypeError('common_queue.dtypes must be [tf.string, tf.string]')
    for dtype in common_queue.dtypes:
      if not dtype.is_compatible_with(tf_dtypes.string):
        raise TypeError('common_queue.dtypes must be [tf.string, tf.string]')

    reader_kwargs = reader_kwargs or {}
    self._readers = [reader_class(**reader_kwargs) for _ in range(num_readers)]
    self._common_queue = common_queue 

def __init__(self,
               reader_class,
               common_queue,
               num_readers=4,
               reader_kwargs=None):
    """ParallelReader creates num_readers instances of the reader_class.

    Each instance is created by calling the `reader_class` function passing
    the arguments specified in `reader_kwargs` as in:
      reader_class(**read_kwargs)

    When you read from a ParallelReader, with its `read()` method,
    you just dequeue examples from the `common_queue`.

    The readers will read different files in parallel, asynchronously enqueueing
    their output into `common_queue`. The `common_queue.dtypes` must be
    [tf.string, tf.string]

    Because each reader can read from a different file, the examples in the
    `common_queue` could be from different files. Due to the asynchronous
    reading there is no guarantee that all the readers will read the same
    number of examples.

    If the `common_queue` is a shuffling queue, then the examples are shuffled.

    Usage:
      common_queue = tf.RandomShuffleQueue(
          capacity=256,
          min_after_dequeue=128,
          dtypes=[tf.string, tf.string])
      p_reader = ParallelReader(tf.TFRecordReader, common_queue)

      common_queue = tf.FIFOQueue(
          capacity=256,
          dtypes=[tf.string, tf.string])
      p_reader = ParallelReader(readers, common_queue, num_readers=2)


    Args:
      reader_class: one of the io_ops.ReaderBase subclasses ex: TFRecordReader
      common_queue: a Queue to hold (key, value pairs) with `dtypes` equal to
        [tf.string, tf.string]. Must be one of the data_flow_ops.Queues
        instances, ex. `tf.FIFOQueue()`, `tf.RandomShuffleQueue()`, ...
      num_readers: a integer, number of instances of reader_class to create.
      reader_kwargs: an optional dict of kwargs to create the readers.

    Raises:
      TypeError: if `common_queue.dtypes` is not [tf.string, tf.string].
    """
    if len(common_queue.dtypes) != 2:
      raise TypeError('common_queue.dtypes must be [tf.string, tf.string]')
    for dtype in common_queue.dtypes:
      if not dtype.is_compatible_with(tf_dtypes.string):
        raise TypeError('common_queue.dtypes must be [tf.string, tf.string]')

    reader_kwargs = reader_kwargs or {}
    self._readers = [reader_class(**reader_kwargs) for _ in range(num_readers)]
    self._common_queue = common_queue 

def __init__(self,
               reader_class,
               common_queue,
               num_readers=4,
               reader_kwargs=None):
    """ParallelReader creates num_readers instances of the reader_class.

    Each instance is created by calling the `reader_class` function passing
    the arguments specified in `reader_kwargs` as in:
      reader_class(**read_kwargs)

    When you read from a ParallelReader, with its `read()` method,
    you just dequeue examples from the `common_queue`.

    The readers will read different files in parallel, asynchronously enqueueing
    their output into `common_queue`. The `common_queue.dtypes` must be
    [tf.string, tf.string]

    Because each reader can read from a different file, the examples in the
    `common_queue` could be from different files. Due to the asynchronous
    reading there is no guarantee that all the readers will read the same
    number of examples.

    If the `common_queue` is a shuffling queue, then the examples are shuffled.

    Usage:
      common_queue = tf.queue.RandomShuffleQueue(
          capacity=256,
          min_after_dequeue=128,
          dtypes=[tf.string, tf.string])
      p_reader = ParallelReader(tf.compat.v1.TFRecordReader, common_queue)

      common_queue = tf.queue.FIFOQueue(
          capacity=256,
          dtypes=[tf.string, tf.string])
      p_reader = ParallelReader(readers, common_queue, num_readers=2)


    Args:
      reader_class: one of the io_ops.ReaderBase subclasses ex: TFRecordReader
      common_queue: a Queue to hold (key, value pairs) with `dtypes` equal to
        [tf.string, tf.string]. Must be one of the data_flow_ops.Queues
        instances, ex. `tf.queue.FIFOQueue()`, `tf.queue.RandomShuffleQueue()`,
        ...
      num_readers: a integer, number of instances of reader_class to create.
      reader_kwargs: an optional dict of kwargs to create the readers.

    Raises:
      TypeError: if `common_queue.dtypes` is not [tf.string, tf.string].
    """
    if len(common_queue.dtypes) != 2:
      raise TypeError('common_queue.dtypes must be [tf.string, tf.string]')
    for dtype in common_queue.dtypes:
      if not dtype.is_compatible_with(tf_dtypes.string):
        raise TypeError('common_queue.dtypes must be [tf.string, tf.string]')

    reader_kwargs = reader_kwargs or {}
    self._readers = [reader_class(**reader_kwargs) for _ in range(num_readers)]
    self._common_queue = common_queue 

def __init__(self,
               reader_class,
               common_queue,
               num_readers=4,
               reader_kwargs=None):
    """ParallelReader creates num_readers instances of the reader_class.

    Each instance is created by calling the `reader_class` function passing
    the arguments specified in `reader_kwargs` as in:
      reader_class(**read_kwargs)

    When you read from a ParallelReader, with its `read()` method,
    you just dequeue examples from the `common_queue`.

    The readers will read different files in parallel, asynchronously enqueueing
    their output into `common_queue`. The `common_queue.dtypes` must be
    [tf.string, tf.string]

    Because each reader can read from a different file, the examples in the
    `common_queue` could be from different files. Due to the asynchronous
    reading there is no guarantee that all the readers will read the same
    number of examples.

    If the `common_queue` is a shuffling queue, then the examples are shuffled.

    Usage:
      common_queue = tf.RandomShuffleQueue(
          capacity=256,
          min_after_dequeue=128,
          dtypes=[tf.string, tf.string])
      p_reader = ParallelReader(tf.TFRecordReader, common_queue)

      common_queue = tf.FIFOQueue(
          capacity=256,
          dtypes=[tf.string, tf.string])
      p_reader = ParallelReader(readers, common_queue, num_readers=2)


    Args:
      reader_class: one of the io_ops.ReaderBase subclasses ex: TFRecordReader
      common_queue: a Queue to hold (key, value pairs) with `dtypes` equal to
        [tf.string, tf.string]. Must be one of the data_flow_ops.Queues
        instances, ex. `tf.FIFOQueue()`, `tf.RandomShuffleQueue()`, ...
      num_readers: a integer, number of instances of reader_class to create.
      reader_kwargs: an optional dict of kwargs to create the readers.

    Raises:
      TypeError: if `common_queue.dtypes` is not [tf.string, tf.string].
    """
    if len(common_queue.dtypes) != 2:
      raise TypeError('common_queue.dtypes must be [tf.string, tf.string]')
    for dtype in common_queue.dtypes:
      if not dtype.is_compatible_with(tf_dtypes.string):
        raise TypeError('common_queue.dtypes must be [tf.string, tf.string]')

    reader_kwargs = reader_kwargs or {}
    self._readers = [reader_class(**reader_kwargs) for _ in range(num_readers)]
    self._common_queue = common_queue 

def __init__(self,
               reader_class,
               common_queue,
               num_readers=4,
               reader_kwargs=None):
    """ParallelReader creates num_readers instances of the reader_class.

    Each instance is created by calling the `reader_class` function passing
    the arguments specified in `reader_kwargs` as in:
      reader_class(**read_kwargs)

    When you read from a ParallelReader, with its `read()` method,
    you just dequeue examples from the `common_queue`.

    The readers will read different files in parallel, asynchronously enqueueing
    their output into `common_queue`. The `common_queue.dtypes` must be
    [tf.string, tf.string]

    Because each reader can read from a different file, the examples in the
    `common_queue` could be from different files. Due to the asynchronous
    reading there is no guarantee that all the readers will read the same
    number of examples.

    If the `common_queue` is a shuffling queue, then the examples are shuffled.

    Usage:
      common_queue = tf.RandomShuffleQueue(
          capacity=256,
          min_after_dequeue=128,
          dtypes=[tf.string, tf.string])
      p_reader = ParallelReader(tf.TFRecordReader, common_queue)

      common_queue = tf.FIFOQueue(
          capacity=256,
          dtypes=[tf.string, tf.string])
      p_reader = ParallelReader(readers, common_queue, num_readers=2)


    Args:
      reader_class: one of the io_ops.ReaderBase subclasses ex: TFRecordReader
      common_queue: a Queue to hold (key, value pairs) with `dtypes` equal to
        [tf.string, tf.string]. Must be one of the data_flow_ops.Queues
        instances, ex. `tf.FIFOQueue()`, `tf.RandomShuffleQueue()`, ...
      num_readers: a integer, number of instances of reader_class to create.
      reader_kwargs: an optional dict of kwargs to create the readers.

    Raises:
      TypeError: if `common_queue.dtypes` is not [tf.string, tf.string].
    """
    if len(common_queue.dtypes) != 2:
      raise TypeError('common_queue.dtypes must be [tf.string, tf.string]')
    for dtype in common_queue.dtypes:
      if not dtype.is_compatible_with(tf_dtypes.string):
        raise TypeError('common_queue.dtypes must be [tf.string, tf.string]')

    reader_kwargs = reader_kwargs or {}
    self._readers = [reader_class(**reader_kwargs) for _ in range(num_readers)]
    self._common_queue = common_queue