Python summary.scalar函数代码示例

def _training_loss(
    features, labels, logits, loss_fn, weight_column_name=None, head_name=None):
  """Returns training loss tensor.

  Training loss is different from the loss reported on the tensorboard as we
  should respect the example weights when computing the gradient.

    L = sum_{i} w_{i} * l_{i} / B

  where B is the number of examples in the batch, l_{i}, w_{i} are individual
  losses, and example weight.

  Args:
    features: Features `dict`.
    labels: Either a `Tensor` for labels or in multihead case, a `dict` of
      string to `Tensor`.
    logits: logits, a float `Tensor`. Shape is `(batch_size, logits_dimension)`.
    loss_fn: Function taking `logits` and `labels`, and returning the raw
      unweighted loss.
    weight_column_name: Key for weights `Tensor` in `features`, if applicable.
    head_name: Head name, used for summary.

  Returns:
    A loss `Output`.
  """
  with ops.name_scope(
      None, "training_loss",
      tuple(six.itervalues(features)) + (labels, logits)) as name:
    loss, weighted_average_loss = _loss(
        loss_fn(logits, labels),
        _weight_tensor(features, weight_column_name),
        name=name)
    summary.scalar(_head_prefixed(head_name, "loss"), weighted_average_loss)
    return loss

  def gradient_clipping(grads_and_vars):
    """Internal function for adaptive clipping."""
    grads, variables = zip(*grads_and_vars)

    norm = clip_ops.global_norm(grads)

    max_norm, log_mean = _adaptive_max_norm(
        norm, std_factor, decay, global_step, epsilon, name)

    # reports the max gradient norm for debugging
    if report_summary:
      summary.scalar("global_norm/adaptive_max_gradient_norm", max_norm)

    # factor will be 1. if norm is smaller than max_norm
    factor = array_ops.where(norm < max_norm,
                             array_ops.ones_like(norm),
                             math_ops.exp(log_mean) / norm)

    if static_max_norm is not None:
      factor = math_ops.minimum(static_max_norm / norm, factor)

    # apply factor
    clipped_grads = []
    for grad in grads:
      if grad is None:
        clipped_grads.append(None)
      elif isinstance(grad, ops.IndexedSlices):
        clipped_grads.append(ops.IndexedSlices(
            grad.values * factor, grad.indices, grad.dense_shape))
      else:
        clipped_grads.append(grad * factor)

    return list(zip(clipped_grads, variables))

 def model_fn_with_summary(features, labels, mode, params):
   del features, labels, params
   loss = constant_op.constant(_EXPECTED_LOSS)
   summary.scalar('loss_scalar_summary', loss)
   summary.histogram('loss_histogram_summary', loss)
   summary.image('loss_image_summary', loss)
   return tpu_estimator.TPUEstimatorSpec(mode=mode, loss=loss)

def _centered_bias(num_label_columns):
  centered_bias = variables.Variable(
      array_ops.zeros([num_label_columns]),
      collections=[_CENTERED_BIAS, ops.GraphKeys.VARIABLES],
      name=_CENTERED_BIAS_WEIGHT)
  summary.scalar(["centered_bias %d" % cb for cb in range(num_label_columns)],
                 array_ops.reshape(centered_bias, [-1]))
  return centered_bias

def _centered_bias(num_label_columns):
  centered_bias = variables.Variable(
      array_ops.zeros([num_label_columns]),
      collections=[_CENTERED_BIAS, ops.GraphKeys.GLOBAL_VARIABLES],
      name=_CENTERED_BIAS_WEIGHT)
  for i in range(num_label_columns):
    summary.scalar("centered_bias %d" % i, centered_bias[i])
  return centered_bias

def prefetch_queue(tensors,
                   capacity=8,
                   num_threads=1,
                   shared_name=None,
                   name=None):
  """Creates a queue to prefetech tensors from `tensors`.

  A queue runner for enqueing tensors into the prefetch_queue is automatically
  added to the TF QueueRunners collection.

  Example:
  This is for example useful to pre-assemble input batches read with
  `tf.train.batch()` and enqueue the pre-assembled batches.  Ops that dequeue
  from the pre-assembled queue will not pay the cost of assembling the batch.

  images, labels = tf.train.batch([image, label], batch_size=32, num_threads=4)
  batch_queue = prefetch_queue([images, labels])
  images, labels = batch_queue.dequeue()
  logits = Net(images)
  loss = Loss(logits, labels)

  Args:
    tensors: A list or dictionary of `Tensors` to enqueue in the buffer.
    capacity: An integer. The maximum number of elements in the queue.
    num_threads: An integer.  Number of threads running the enqueue op.
    shared_name: (optional). If set, this queue will be shared under the given
      name across multiple sessions.
    name: (Optional) A name for the operations.

  Returns:
    A queue from which you can dequeue tensors with the same type and shape
    as `tensors`.
  """
  if isinstance(tensors, dict):
    # Need to wrap the keys and values in list() since Python3 returns views.
    # We sort the keys so the order is consistent across runs.
    names = list(sorted(tensors.keys()))
    tensor_list = list([tensors[n] for n in names])
  else:
    names = None
    tensor_list = tensors

  with ops.name_scope(name, "prefetch_queue", tensor_list) as name:
    dtypes = [t.dtype for t in tensor_list]
    shapes = [t.get_shape() for t in tensor_list]
    queue = data_flow_ops.FIFOQueue(capacity=capacity,
                                    dtypes=dtypes,
                                    shapes=shapes,
                                    names=names,
                                    shared_name=shared_name)
    enqueue_op = queue.enqueue(tensors)
    queue_runner.add_queue_runner(
        queue_runner.QueueRunner(queue, [enqueue_op] * num_threads))
    summary.scalar("fraction_of_%d_full" % capacity,
                   math_ops.to_float(queue.size()) * (1. / capacity))
    return queue

def input_producer(input_tensor, element_shape=None, num_epochs=None,
                   shuffle=True, seed=None, capacity=32, shared_name=None,
                   summary_name=None, name=None):
  """Output the rows of `input_tensor` to a queue for an input pipeline.

  Args:
    input_tensor: A tensor with the rows to produce. Must be at least
      one-dimensional. Must either have a fully-defined shape, or
      `element_shape` must be defined.
    element_shape: (Optional.) A `TensorShape` representing the shape of a
      row of `input_tensor`, if it cannot be inferred.
    num_epochs: (Optional.) An integer. If specified `input_producer` produces
      each row of `input_tensor` `num_epochs` times before generating an
      `OutOfRange` error. If not specified, `input_producer` can cycle through
      the rows of `input_tensor` an unlimited number of times.
    shuffle: (Optional.) A boolean. If true, the rows are randomly shuffled
      within each epoch.
    seed: (Optional.) An integer. The seed to use if `shuffle` is true.
    capacity: (Optional.) The capacity of the queue to be used for buffering
      the input.
    shared_name: (Optional.) If set, this queue will be shared under the given
      name across multiple sessions.
    summary_name: (Optional.) If set, a scalar summary for the current queue
      size will be generated, using this name as part of the tag.
    name: (Optional.) A name for queue.

  Returns:
    A queue with the output rows.  A `QueueRunner` for the queue is
    added to the current `QUEUE_RUNNER` collection of the current
    graph.

  Raises:
    ValueError: If the shape of the input cannot be inferred from the arguments.
  """
  with ops.name_scope(name, "input_producer", [input_tensor]):
    input_tensor = ops.convert_to_tensor(input_tensor, name="input_tensor")
    element_shape = input_tensor.get_shape()[1:].merge_with(element_shape)
    if not element_shape.is_fully_defined():
      raise ValueError("Either `input_tensor` must have a fully defined shape "
                       "or `element_shape` must be specified")

    if shuffle:
      input_tensor = random_ops.random_shuffle(input_tensor, seed=seed)

    input_tensor = limit_epochs(input_tensor, num_epochs)

    q = data_flow_ops.FIFOQueue(capacity=capacity,
                                dtypes=[input_tensor.dtype.base_dtype],
                                shapes=[element_shape],
                                shared_name=shared_name, name=name)
    enq = q.enqueue_many([input_tensor])
    queue_runner.add_queue_runner(queue_runner.QueueRunner(q, [enq]))
    if summary_name is not None:
      summary.scalar("queue/%s/%s" % (q.name, summary_name),
                     math_ops.cast(q.size(), dtypes.float32) * (1. / capacity))
    return q

def _centered_bias(logits_dimension, weight_collection):
  """Creates and returns centered bias."""
  centered_bias = variables.Variable(
      array_ops.zeros([logits_dimension]),
      collections=[weight_collection, ops.GraphKeys.GLOBAL_VARIABLES],
      name="centered_bias_weight")

  biases = array_ops.reshape(centered_bias, [-1])
  for cb in range(logits_dimension):
    summary.scalar("centered_bias_%d" % cb, biases[cb])
  return centered_bias

def _conditional_batch(tensors, keep_input, batch_size, num_threads=10):
  """Conditionally enqueue tensors based on accept_prob.

  Specifically, enqueue the element if accept_prob > rand_unif([0, 1]).

  Args:
      tensors: List of tensors to enqueue.
      keep_input: Bool. Whether to enqueue or not.
      batch_size: Size of batch.
      num_threads: Number of enqueueing threads.

  Returns:
      List of batched tensors.

  Raises:
      ValueError: `accept_prob` isn't 0D.
  """
  keep_input.get_shape().assert_has_rank(0)
  # Determine shapes and types of to-be-enqueued-tensors.
  shapes_list = []
  dtypes_list = []
  for tensor in tensors:
    cur_shape = tensor.get_shape()
    cur_shape.assert_is_fully_defined()
    shapes_list.append(cur_shape)
    dtypes_list.append(tensor.dtype)

  final_q = data_flow_ops.FIFOQueue(capacity=batch_size,
                                    shapes=shapes_list,
                                    dtypes=dtypes_list,
                                    name='batched_queue')
  summary.scalar('queue/%s/size' % final_q.name, final_q.size())

  # Conditionally enqueue.
  # Reshape enqueue op to match no_op's shape.
  conditional_enqueue = control_flow_ops.cond(
      keep_input,
      lambda: final_q.enqueue(tensors),
      control_flow_ops.no_op)
  queue_runner.add_queue_runner(queue_runner.QueueRunner(
      final_q, [conditional_enqueue] * num_threads))

  out_tensor = final_q.dequeue_many(batch_size)
  # Queues return a single tensor if the list of enqued tensors is one. Since we
  # want the type to be the same in all cases, always return a list.
  if isinstance(out_tensor, ops.Tensor):
    out_tensor = [out_tensor]

  return out_tensor

  def test_report_unsupported_operations(self):
    """Tests that unsupported operations are detected."""
    context = self.create_test_xla_compile_context()
    context.Enter()
    dummy_tensor = constant_op.constant(1.1)
    audio_summary = summary.audio('audio_summary', dummy_tensor, 0.5)
    histogram_summary = summary.histogram('histogram_summary', dummy_tensor)
    image_summary = summary.image('image_summary', dummy_tensor)
    scalar_summary = summary.scalar('scalar_summary', dummy_tensor)
    tensor_summary = summary.tensor_summary('tensor_summary', dummy_tensor)
    summary.merge(
        [
            audio_summary, histogram_summary, image_summary, scalar_summary,
            tensor_summary
        ],
        name='merge_summary')
    logging_ops.Print(dummy_tensor, [dummy_tensor], name='print_op')
    context.Exit()

    unsupported_ops_names = [op.name for op in context._unsupported_ops]
    self.assertEqual(unsupported_ops_names, [
        u'audio_summary', u'histogram_summary', u'image_summary',
        u'scalar_summary', u'tensor_summary', u'merge_summary/merge_summary',
        u'print_op'
    ])

  def __init__(self,
               examples,
               variables,
               options):
    """Create a new sdca optimizer."""

    if not examples or not variables or not options:
      raise ValueError('examples, variables and options must all be specified.')

    supported_losses = ('logistic_loss', 'squared_loss', 'hinge_loss',
                        'smooth_hinge_loss')
    if options['loss_type'] not in supported_losses:
      raise ValueError('Unsupported loss_type: ', options['loss_type'])

    self._assertSpecified(['example_labels', 'example_weights', 'example_ids',
                           'sparse_features', 'dense_features'], examples)
    self._assertList(['sparse_features', 'dense_features'], examples)

    self._assertSpecified(['sparse_features_weights', 'dense_features_weights'],
                          variables)
    self._assertList(['sparse_features_weights', 'dense_features_weights'],
                     variables)

    self._assertSpecified(['loss_type', 'symmetric_l2_regularization',
                           'symmetric_l1_regularization'], options)

    for name in ['symmetric_l1_regularization', 'symmetric_l2_regularization']:
      value = options[name]
      if value < 0.0:
        raise ValueError('%s should be non-negative. Found (%f)' %
                         (name, value))

    self._examples = examples
    self._variables = variables
    self._options = options
    self._create_slots()
    self._hashtable = ShardedMutableDenseHashTable(
        key_dtype=dtypes.int64,
        value_dtype=dtypes.float32,
        num_shards=self._num_table_shards(),
        default_value=[0.0, 0.0, 0.0, 0.0],
        # SdcaFprint never returns 0 or 1 for the low64 bits, so this a safe
        # empty_key (that will never collide with actual payloads).
        empty_key=[0, 0])

    summary.scalar('approximate_duality_gap', self.approximate_duality_gap())
    summary.scalar('examples_seen', self._hashtable.size())

  def _training_loss(self, features, labels, logits=None,
                     logits_input=None, name="training_loss"):
    """Returns training loss tensor for this head.

    Training loss is different from the loss reported on the tensorboard as we
    should respect the example weights when computing the gradient.

      L = sum_{i} w_{i} * l_{i} / B

    where B is the number of examples in the batch, l_{i}, w_{i} are individual
    losses, and example weight.

    Args:
      features: features dict.
      labels: either a tensor for labels or in multihead case, a dict of string
        to labels tensor.
      logits: logits, a float tensor.
      logits_input: Output of last hidden layer.
      name: Op name.

    Returns:
      A tuple of training Loss and additional_train_op (possibly None)
    """
    labels = _check_labels(labels, self._label_name)

    centered_bias_step = None
    if self._enable_centered_bias:
      logits = nn.bias_add(logits, _centered_bias(
          self.logits_dimension,
          self._centered_bias_weight_collection))
      centered_bias_step = [_centered_bias_step(
          self.logits_dimension,
          self._centered_bias_weight_collection,
          labels,
          self._train_loss_fn)]

    loss_unweighted = self._train_loss_fn(logits, labels)
    loss, weighted_average_loss = _loss(
        loss_unweighted,
        _weight_tensor(features, self._weight_column_name),
        name=name)
    summary.scalar(
        _head_prefixed(self._head_name, "loss"), weighted_average_loss)
    return loss, centered_bias_step

def _centered_bias(logits_dimension):
  """Returns `logits`, optionally with centered bias applied.

  Args:
    logits_dimension: Last dimension of `logits`. Must be >= 1.

  Returns:
    Centered bias `Variable`.

  Raises:
    ValueError: if `logits_dimension` is invalid.
  """
  if (logits_dimension is None) or (logits_dimension < 1):
    raise ValueError("Invalid logits_dimension %s." % logits_dimension)
  centered_bias = variable_scope.get_variable(
      name="centered_bias_weight",
      shape=(logits_dimension,),
      initializer=init_ops.zeros_initializer,
      trainable=True)
  for dim in range(logits_dimension):
    summary.scalar("centered_bias_%d" % dim, centered_bias[dim])
  return centered_bias

def _add_scalar_summary(tensor, tag=None):
    """Add a scalar summary operation for the tensor.

  Args:
    tensor: The tensor to summarize.
    tag: The tag to use, if None then use tensor's op's name.

  Returns:
    The created histogram summary.

  Raises:
    ValueError: If the tag is already in use or the rank is not 0.
  """
    tensor.get_shape().assert_has_rank(0)
    tag = tag or "%s_summary" % tensor.op.name
    return summary.scalar(tag, tensor)

def batch_join(tensors_list, batch_size, capacity=32, enqueue_many=False,
               shapes=None, dynamic_pad=False, allow_smaller_final_batch=False,
               shared_name=None, name=None):
  """Runs a list of tensors to fill a queue to create batches of examples.

  The `tensors_list` argument is a list of tuples of tensors, or a list of
  dictionaries of tensors.  Each element in the list is treated similarly
  to the `tensors` argument of `tf.train.batch()`.

  Enqueues a different list of tensors in different threads.
  Implemented using a queue -- a `QueueRunner` for the queue
  is added to the current `Graph`'s `QUEUE_RUNNER` collection.

  `len(tensors_list)` threads will be started,
  with thread `i` enqueuing the tensors from
  `tensors_list[i]`. `tensors_list[i1][j]` must match
  `tensors_list[i2][j]` in type and shape, except in the first
  dimension if `enqueue_many` is true.

  If `enqueue_many` is `False`, each `tensors_list[i]` is assumed
  to represent a single example. An input tensor `x` will be output as a
  tensor with shape `[batch_size] + x.shape`.

  If `enqueue_many` is `True`, `tensors_list[i]` is assumed to
  represent a batch of examples, where the first dimension is indexed
  by example, and all members of `tensors_list[i]` should have the
  same size in the first dimension.  The slices of any input tensor
  `x` are treated as examples, and the output tensors will have shape
  `[batch_size] + x.shape[1:]`.

  The `capacity` argument controls the how long the prefetching is allowed to
  grow the queues.

  The returned operation is a dequeue operation and will throw
  `tf.errors.OutOfRangeError` if the input queue is exhausted. If this
  operation is feeding another input queue, its queue runner will catch
  this exception, however, if this operation is used in your main thread
  you are responsible for catching this yourself.

  *N.B.:* If `dynamic_pad` is `False`, you must ensure that either
  (i) the `shapes` argument is passed, or (ii) all of the tensors in
  `tensors_list` must have fully-defined shapes. `ValueError` will be
  raised if neither of these conditions holds.

  If `dynamic_pad` is `True`, it is sufficient that the *rank* of the
  tensors is known, but individual dimensions may have value `None`.
  In this case, for each enqueue the dimensions with value `None`
  may have a variable length; upon dequeue, the output tensors will be padded
  on the right to the maximum shape of the tensors in the current minibatch.
  For numbers, this padding takes value 0.  For strings, this padding is
  the empty string.  See `PaddingFIFOQueue` for more info.

  If `allow_smaller_final_batch` is `True`, a smaller batch value than
  `batch_size` is returned when the queue is closed and there are not enough
  elements to fill the batch, otherwise the pending elements are discarded.
  In addition, all output tensors' static shapes, as accessed via the
  `get_shape` method will have a first `Dimension` value of `None`, and
  operations that depend on fixed batch_size would fail.

  Args:
    tensors_list: A list of tuples or dictionaries of tensors to enqueue.
    batch_size: An integer. The new batch size pulled from the queue.
    capacity: An integer. The maximum number of elements in the queue.
    enqueue_many: Whether each tensor in `tensor_list_list` is a single
      example.
    shapes: (Optional) The shapes for each example.  Defaults to the
      inferred shapes for `tensor_list_list[i]`.
    dynamic_pad: Boolean.  Allow variable dimensions in input shapes.
      The given dimensions are padded upon dequeue so that tensors within a
      batch have the same shapes.
    allow_smaller_final_batch: (Optional) Boolean. If `True`, allow the final
      batch to be smaller if there are insufficient items left in the queue.
    shared_name: (Optional) If set, this queue will be shared under the given
      name across multiple sessions.
    name: (Optional) A name for the operations.

  Returns:
    A list or dictionary of tensors with the same number and types as
    `tensors_list[i]`.

  Raises:
    ValueError: If the `shapes` are not specified, and cannot be
      inferred from the elements of `tensor_list_list`.
  """
  tensor_list_list = _as_tensor_list_list(tensors_list)
  with ops.name_scope(name, "batch_join", _flatten(tensor_list_list)) as name:
    tensor_list_list = _validate_join(tensor_list_list)
    tensor_list_list, sparse_info = _store_sparse_tensors_join(
        tensor_list_list, enqueue_many)
    types = _dtypes(tensor_list_list)
    shapes = _shapes(tensor_list_list, shapes, enqueue_many)
    # TODO(josh11b,mrry): Switch to BatchQueue once it is written.
    queue = _which_queue(dynamic_pad)(
        capacity=capacity, dtypes=types, shapes=shapes, shared_name=shared_name)
    _enqueue_join(queue, tensor_list_list, enqueue_many)
    summary.scalar("queue/%s/fraction_of_%d_full" % (queue.name, capacity),
                   math_ops.cast(queue.size(), dtypes.float32) *
                   (1. / capacity))

    if allow_smaller_final_batch:
#.........这里部分代码省略.........