Python tensorflow.python.framework.ops.get_default_graph() Examples

def _init_from_proto(self, variable_def, import_scope=None):
    """Creates a new variable from `VariableDef` protocol buffer.

    Args:
      variable_def: `VariableDef` protocol buffer.
      import_scope: Optional `string`. Name scope to add.
    """
    assert isinstance(variable_def, variable_pb2.VariableDef)
    # Create from variable_def.
    g = ops.get_default_graph()
    self._variable = g.as_graph_element(
        ops.prepend_name_scope(variable_def.variable_name,
                               import_scope=import_scope))
    self._initializer_op = g.as_graph_element(
        ops.prepend_name_scope(variable_def.initializer_name,
                               import_scope=import_scope))
    self._snapshot = g.as_graph_element(
        ops.prepend_name_scope(variable_def.snapshot_name,
                               import_scope=import_scope))
    if variable_def.HasField("save_slice_info_def"):
      self._save_slice_info = Variable.SaveSliceInfo(
          save_slice_info_def=variable_def.save_slice_info_def)
    else:
      self._save_slice_info = None
    self._caching_device = None 

def convert_collection_to_dict(collection, clear_collection=False):
  """Returns an OrderedDict of Tensors with their aliases as keys.

  Args:
    collection: A collection.
    clear_collection: When True, it clears the collection after converting to
      OrderedDict.

  Returns:
    An OrderedDict of {alias: tensor}
  """
  output = OrderedDict((alias, tensor)
                       for tensor in ops.get_collection(collection)
                       for alias in get_tensor_aliases(tensor))
  if clear_collection:
    ops.get_default_graph().clear_collection(collection)
  return output 

def _lower_bound(inputs, bound, name=None):
    """Same as tf.maximum, but with helpful gradient for inputs < bound.

    The gradient is overwritten so that it is passed through if the input is not
    hitting the bound. If it is, only gradients that push `inputs` higher than
    the bound are passed through. No gradients are passed through to the bound.

    Args:
      inputs: input tensor
      bound: lower bound for the input tensor
      name: name for this op

    Returns:
      tf.maximum(inputs, bound)
    """
    with ops.name_scope(name, 'GDNLowerBound', [inputs, bound]) as scope:
      inputs = ops.convert_to_tensor(inputs, name='inputs')
      bound = ops.convert_to_tensor(bound, name='bound')
      with ops.get_default_graph().gradient_override_map(
          {'Maximum': 'GDNLowerBound'}):
        return math_ops.maximum(inputs, bound, name=scope) 

def _init_from_proto(self, context_def, import_scope=None):
    """Creates a new `CondContext` from protocol buffer.

    Args:
      context_def: `CondContextDef` protocol buffer.
      import_scope: Optional `string`. Name scope to add.
    """
    assert isinstance(context_def, control_flow_pb2.CondContextDef)
    # Create from context_def.
    g = ops.get_default_graph()
    self._name = ops.prepend_name_scope(
        context_def.context_name, import_scope)
    self._pred = g.as_graph_element(ops.prepend_name_scope(
        context_def.pred_name, import_scope))
    self._pivot = g.as_graph_element(ops.prepend_name_scope(
        context_def.pivot_name, import_scope))
    self._branch = context_def.branch
    super(CondContext, self).__init__(values_def=context_def.values_def,
                                      import_scope=import_scope) 

def AddOp(self, op):
    """Add `op` to the current context."""
    # For a reduction op, if op is in a grad context and its input is from
    # its forward context, moving op to the forward context means we would
    # store the tensor after the reduction as opposed to the tensor before
    # reduction, and therefore could significantly reduce memory consumption.
    # For now, we do this only for a few ops.
    if op.type in {"Shape", "Size", "Rank"}:
      grad_ctxt = ops.get_default_graph()._get_control_flow_context()
      if grad_ctxt:
        grad_ctxt = grad_ctxt.GetWhileContext()
        if grad_ctxt.grad_state:
          op_input_forward_ctxt = _GetWhileContext(op.inputs[0].op)
          if op_input_forward_ctxt == grad_ctxt.grad_state.forward_context:
            op_input_ctxt = op.inputs[0].op._get_control_flow_context()
            op._set_control_flow_context(op_input_ctxt)
            op_input_ctxt._AddOpInternal(op)
            return
    self._AddOpInternal(op) 

def _FixControlInputsAndContext(self, enters):
    graph = ops.get_default_graph()
    # pylint: disable=protected-access
    for e in enters:
      if isinstance(e, ops.Tensor):
        xs = [e]
      else:
        if not isinstance(e, (ops.IndexedSlices, sparse_tensor.SparseTensor)):
          raise TypeError("Type %s not supported" % type(e))
        xs = [e.values, e.indices]
        shape = e.dense_shape
        if shape is not None:
          xs.append(shape)
      for x in xs:
        inp_op = x.op.inputs[0]
        control_inputs = graph._control_dependencies_for_inputs([inp_op])
        outer_control_inputs = [op for op in control_inputs
                                if self._IsInOuterContext(op)]
        x.op._set_control_flow_context(self)
        x.op._add_control_inputs(outer_control_inputs)
        graph._record_op_seen_by_control_dependencies(x.op)
    # pylint: enable=protected-access 

def add_check_numerics_ops():
  """Connect a `check_numerics` to every floating point tensor.

  `check_numerics` operations themselves are added for each `half`, `float`,
  or `double` tensor in the graph. For all ops in the graph, the
  `check_numerics` op for all of its (`half`, `float`, or `double`) inputs
  is guaranteed to run before the `check_numerics` op on any of its outputs.

  Returns:
    A `group` op depending on all `check_numerics` ops added.
  """
  check_op = []
  # This code relies on the ordering of ops in get_operations().
  # The producer of a tensor always comes before that tensor's consumer in
  # this list. This is true because get_operations() returns ops in the order
  # added, and an op can only be added after its inputs are added.
  for op in ops.get_default_graph().get_operations():
    for output in op.outputs:
      if output.dtype in [dtypes.float16, dtypes.float32, dtypes.float64]:
        message = op.name + ":" + str(output.value_index)
        with ops.control_dependencies(check_op):
          check_op = [array_ops.check_numerics(output, message=message)]
  return control_flow_ops.group(*check_op) 

def _get_or_create_eval_step():
  """Gets or creates the eval step `Tensor`.

  Returns:
    A `Tensor` representing a counter for the evaluation step.

  Raises:
    ValueError: If multiple `Tensors` have been added to the
      `tf.GraphKeys.EVAL_STEP` collection.
  """
  graph = ops.get_default_graph()
  eval_steps = graph.get_collection(ops.GraphKeys.EVAL_STEP)
  if len(eval_steps) == 1:
    return eval_steps[0]
  elif len(eval_steps) > 1:
    raise ValueError('Multiple tensors added to tf.GraphKeys.EVAL_STEP')
  else:
    counter = variable_scope.get_variable(
        'eval_step',
        shape=[],
        dtype=dtypes.int64,
        initializer=init_ops.zeros_initializer(),
        trainable=False,
        collections=[ops.GraphKeys.LOCAL_VARIABLES, ops.GraphKeys.EVAL_STEP])
    return counter 

def before_run(self, run_context):  # pylint: disable=unused-argument
    if self._timer.last_triggered_step() is None:
      # We do write graph and saver_def at the first call of before_run.
      # We cannot do this in begin, since we let other hooks to change graph and
      # add variables in begin. Graph is finalized after all begin calls.
      training_util.write_graph(
          ops.get_default_graph().as_graph_def(add_shapes=True),
          self._checkpoint_dir,
          "graph.pbtxt")
      saver_def = self._get_saver().saver_def if self._get_saver() else None
      graph = ops.get_default_graph()
      meta_graph_def = meta_graph.create_meta_graph_def(
          graph_def=graph.as_graph_def(add_shapes=True),
          saver_def=saver_def)
      self._summary_writer.add_graph(graph)
      self._summary_writer.add_meta_graph(meta_graph_def)

    return SessionRunArgs(self._global_step_tensor) 

def _as_graph_element(obj):
  """Retrieves Graph element."""
  graph = ops.get_default_graph()
  if not isinstance(obj, six.string_types):
    if not hasattr(obj, "graph") or obj.graph != graph:
      raise ValueError("Passed %s should have graph attribute that is equal "
                       "to current graph %s." % (obj, graph))
    return obj
  if ":" in obj:
    element = graph.as_graph_element(obj)
  else:
    element = graph.as_graph_element(obj + ":0")
    # Check that there is no :1 (e.g. it's single output).
    try:
      graph.as_graph_element(obj + ":1")
    except (KeyError, ValueError):
      pass
    else:
      raise ValueError("Name %s is ambiguous, "
                       "as this `Operation` has multiple outputs "
                       "(at least 2)." % obj)
  return element 

def _init_from_proto(self, queue_runner_def, import_scope=None):
    """Create a QueueRunner from `QueueRunnerDef`.

    Args:
      queue_runner_def: Optional `QueueRunnerDef` protocol buffer.
      import_scope: Optional `string`. Name scope to add.
    """
    assert isinstance(queue_runner_def, queue_runner_pb2.QueueRunnerDef)
    g = ops.get_default_graph()
    self._queue = g.as_graph_element(
        ops.prepend_name_scope(queue_runner_def.queue_name, import_scope))
    self._enqueue_ops = [g.as_graph_element(
        ops.prepend_name_scope(op, import_scope))
                         for op in queue_runner_def.enqueue_op_name]
    self._close_op = g.as_graph_element(ops.prepend_name_scope(
        queue_runner_def.close_op_name, import_scope))
    self._cancel_op = g.as_graph_element(ops.prepend_name_scope(
        queue_runner_def.cancel_op_name, import_scope))
    self._queue_closed_exception_types = tuple(
        errors.exception_type_from_error_code(code)
        for code in queue_runner_def.queue_closed_exception_types)
    # Legacy support for old QueueRunnerDefs created before this field
    # was added.
    if not self._queue_closed_exception_types:
      self._queue_closed_exception_types = (errors.OutOfRangeError,) 

def create_global_step(graph=None):
  """Create global step tensor in graph.

  Args:
    graph: The graph in which to create the global step tensor. If missing,
      use default graph.

  Returns:
    Global step tensor.

  Raises:
    ValueError: if global step tensor is already defined.
  """
  graph = graph or ops.get_default_graph()
  if get_global_step(graph) is not None:
    raise ValueError('"global_step" already exists.')
  # Create in proper graph and base name_scope.
  with graph.as_default() as g, g.name_scope(None):
    return variable_scope.get_variable(
        ops.GraphKeys.GLOBAL_STEP,
        shape=[],
        dtype=dtypes.int64,
        initializer=init_ops.zeros_initializer(),
        trainable=False,
        collections=[ops.GraphKeys.GLOBAL_VARIABLES, ops.GraphKeys.GLOBAL_STEP]) 

def _unique_layer_name(name):
  """Makes a layer name (or arbitrary string) unique within a TensorFlow graph.

  Arguments:
    name: String name to make unique.

  Returns:
    Unique string name.

  Example:

  ```
    >>> _unique_layer_name('dense')
    dense_1
    >>> _unique_layer_name('dense')
    dense_2
  ```
  """
  graph = ops.get_default_graph()
  layer_name_uids = PER_GRAPH_LAYER_NAME_UIDS[graph]
  layer_name_uids[name] += 1
  return name + '_' + str(layer_name_uids[name]) 

def make_one_shot_iterator(self):
    """Creates an `Iterator` for enumerating the elements of this dataset.

    **N.B.** The returned iterator will be initialized automatically.
    A "one-shot" iterator does not currently support re-initialization.

    Returns:
      An `Iterator` over the elements of this dataset.
    """
    # NOTE(mrry): We capture by value here to ensure that `_make_dataset()` is
    # a 0-argument function.
    @function.Defun(capture_by_value=True)
    def _make_dataset():
      return self.make_dataset_resource()

    _make_dataset.add_to_graph(ops.get_default_graph())

    return Iterator(
        gen_dataset_ops.one_shot_iterator(
            dataset_factory=_make_dataset,
            output_types=nest.flatten(self.output_types),
            output_shapes=nest.flatten(self.output_shapes)), None,
        self.output_types, self.output_shapes) 

def _as_graph_element(obj):
  """Retrieves Graph element."""
  graph = ops.get_default_graph()
  if not isinstance(obj, six.string_types):
    if not hasattr(obj, "graph") or obj.graph != graph:
      raise ValueError("Passed %s should have graph attribute that is equal "
                       "to current graph %s." % (obj, graph))
    return obj
  if ":" in obj:
    element = graph.as_graph_element(obj)
  else:
    element = graph.as_graph_element(obj + ":0")
    # Check that there is no :1 (e.g. it's single output).
    try:
      graph.as_graph_element(obj + ":1")
    except (KeyError, ValueError):
      pass
    else:
      raise ValueError("Name %s is ambiguous, "
                       "as this `Operation` has multiple outputs "
                       "(at least 2)." % obj)
  return element 

def get_name_scope():
  """Returns the current name scope of the default graph.

  For example:

    ```python
    with tf.name_scope('scope1'):
      with tf.name_scope('scope2'):
        print(tf.contrib.framework.get_name_scope())
    ```
    would print the string `scope1/scope2`.

  Returns:
    A string represnting the current name scope.
  """
  return ops.get_default_graph().get_name_scope() 

def get_extra_vars():
  """Returns the captured variables by the function.

  Returns:
    If the default graph is being used to define a function, the
    returned list of variables are those created inside the function
    body so far. Otherwise, returns an empty list.
  """
  g = ops.get_default_graph()
  if isinstance(g, _FuncGraph):
    return g.extra_vars
  else:
    return [] 

def _create_slots(self, var_list):
        first_var = min(var_list, key=lambda x: x.name)

        graph = None if context.executing_eagerly() else ops.get_default_graph()
        # Create slots for the first and second moments.
        for v in var_list :
            self._zeros_slot(v, "m", self._name)
            self._zeros_slot(v, "v", self._name)
            self._zeros_slot(v, "vhat", self._name) 

def get_seed(op_seed):
  """Returns the local seeds an operation should use given an op-specific seed.

  Given operation-specific seed, `op_seed`, this helper function returns two
  seeds derived from graph-level and op-level seeds. Many random operations
  internally use the two seeds to allow user to change the seed globally for a
  graph, or for only specific operations.

  For details on how the graph-level seed interacts with op seeds, see
  @{tf.set_random_seed}.

  Args:
    op_seed: integer.

  Returns:
    A tuple of two integers that should be used for the local seed of this
    operation.
  """
  graph_seed = ops.get_default_graph().seed
  if graph_seed is not None:
    if op_seed is None:
      # pylint: disable=protected-access
      op_seed = ops.get_default_graph()._last_id
    seeds = _truncate_seed(graph_seed), _truncate_seed(op_seed)
  else:
    if op_seed is not None:
      seeds = DEFAULT_GRAPH_SEED, _truncate_seed(op_seed)
    else:
      seeds = None, None
  # Avoid (0, 0) as the C++ ops interpret it as nondeterminism, which would
  # be unexpected since Python docs say nondeterminism is (None, None).
  if seeds == (0, 0):
    return (0, _MAXINT32)
  return seeds 

def setUp(self):
    self._ClearCachedSession()
    random.seed(random_seed.DEFAULT_GRAPH_SEED)
    np.random.seed(random_seed.DEFAULT_GRAPH_SEED)
    ops.reset_default_graph()
    ops.get_default_graph().seed = random_seed.DEFAULT_GRAPH_SEED 

def get(logdir):
    """Returns the FileWriter for the specified directory.

    Args:
      logdir: str, name of the directory.

    Returns:
      A `FileWriter`.
    """
    with FileWriterCache._lock:
      if logdir not in FileWriterCache._cache:
        FileWriterCache._cache[logdir] = FileWriter(
            logdir, graph=ops.get_default_graph())
      return FileWriterCache._cache[logdir] 

def get_extra_inputs():
  """Returns the captured input tensors by the function.

  Returns:
    If the default graph is being used to define a function, the
    returned list of tensors are those accessed inside the function body
    but defined outside the function body so far. Otherwise, returns an
    empty list.
  """
  g = ops.get_default_graph()
  if isinstance(g, _FuncGraph):
    return g.extra_inputs
  else:
    return [] 

def copy_scoped_meta_graph(from_scope, to_scope,
                           from_graph=None, to_graph=None):
  """Copies a sub-meta_graph from one scope to another.

  Args:
    from_scope: `String` name scope containing the subgraph to be copied.
    to_scope: `String` name scope under which the copied subgraph will reside.
    from_graph: Optional `Graph` from which to copy the subgraph. If `None`, the
      default graph is use.
    to_graph: Optional `Graph` to which to copy the subgraph. If `None`, the
      default graph is used.

  Returns:
    A dictionary of `Variables` that has been copied into `to_scope`.

  Raises:
    ValueError: If `from_scope` and `to_scope` are the same while
      `from_graph` and `to_graph` are also the same.
  """
  from_graph = from_graph or ops.get_default_graph()
  to_graph = to_graph or ops.get_default_graph()

  if from_graph == to_graph and from_scope == to_scope:
    raise ValueError("'from_scope' and 'to_scope' need to be different "
                     "when performing copy in the same graph.")

  orig_meta_graph, var_list = export_scoped_meta_graph(
      export_scope=from_scope, graph=from_graph)
  var_list = import_scoped_meta_graph(orig_meta_graph,
                                      graph=to_graph,
                                      import_scope=to_scope)
  return var_list 

def __call__(self, *args, **kwargs):
    self.add_to_graph(ops.get_default_graph())
    args = [ops.convert_to_tensor(_) for _ in args] + self._extra_inputs
    ret, op = _call(self._definition.signature, *args, **kwargs)
    if self._shape_func is not None:
      shapes = self._shape_func(op)
      if len(shapes) != len(op.outputs):
        raise ValueError("shape_func produced %d shapes for %d outputs" %
                         (len(shapes), len(op.outputs)))
      for (t, shape) in zip(op.outputs, shapes):
        t.set_shape(shape)
    return ret 

def __init__(self, layers=None, name=None):
    self.layers = []  # Stack of layers.
    self.model = None  # Internal Model instance.
    self.inputs = []  # List of input tensors
    self.outputs = []  # List of length 1: the output tensor (unique).
    self._trainable = True
    self._initial_weights = None

    # Model attributes.
    self.inbound_nodes = []
    self.outbound_nodes = []
    self.built = False

    # Set model name.
    if not name:
      prefix = 'sequential_'
      name = prefix + str(K.get_uid(prefix))
    self.name = name

    # The following properties are not actually used by Keras;
    # they exist for compatibility with TF's variable scoping mechanism.
    self._updates = []
    self._scope = None
    self._reuse = None
    self._base_name = name
    self._graph = ops.get_default_graph()

    # Add to the model any layers passed to the constructor.
    if layers:
      for layer in layers:
        self.add(layer) 

def clear_session():
  """Destroys the current TF graph and creates a new one.

  Useful to avoid clutter from old models / layers.
  """
  global _SESSION
  global _GRAPH_LEARNING_PHASES  # pylint: disable=global-variable-not-assigned
  ops.reset_default_graph()
  reset_uids()
  _SESSION = None
  phase = array_ops.placeholder(dtype='bool', name='keras_learning_phase')
  _GRAPH_LEARNING_PHASES = {}
  _GRAPH_LEARNING_PHASES[ops.get_default_graph()] = phase 

def learning_phase():
  """Returns the learning phase flag.

  The learning phase flag is a bool tensor (0 = test, 1 = train)
  to be passed as input to any Keras function
  that uses a different behavior at train time and test time.

  Returns:
      Learning phase (scalar integer tensor or Python integer).
  """
  graph = ops.get_default_graph()
  if graph not in _GRAPH_LEARNING_PHASES:
    phase = array_ops.placeholder(dtype='bool', name='keras_learning_phase')
    _GRAPH_LEARNING_PHASES[graph] = phase
  return _GRAPH_LEARNING_PHASES[graph] 

def set_learning_phase(value):
  """Sets the learning phase to a fixed value.

  Arguments:
      value: Learning phase value, either 0 or 1 (integers).

  Raises:
      ValueError: if `value` is neither `0` nor `1`.
  """
  global _GRAPH_LEARNING_PHASES  # pylint: disable=global-variable-not-assigned
  if value not in {0, 1}:
    raise ValueError('Expected learning phase to be ' '0 or 1.')
  _GRAPH_LEARNING_PHASES[ops.get_default_graph()] = value 

def __getattribute__(self, name):
    """Forwards to the methods in the current graph's `Operation` object."""
    op_name = object.__getattribute__(self, '_name')
    graph = ops.get_default_graph()

    # Short-circuit getting some of these attributes that are readily
    # available without forwarding to the actual operation. This is done
    # because `get_operation_by_name` tries to acquire the parent graph's
    # lock protecting the nodes_by_* data structures, and these attributes
    # (not requiring the lock) could be queried by other function holding
    # the lock.
    if name == 'name':
      return op_name
    elif name == '_as_graph_element':
      return lambda: self
    elif name == '__class__':
      return OperationProxy
    elif name == 'graph':
      original_graph = object.__getattribute__(self, '_original_graph')
      if original_graph.is_child_graph(graph):
        return graph
      else:
        return original_graph
    else:
      op = graph.get_operation_by_name(op_name)
      return getattr(op, name) 

def get_extra_args():
  """Returns the corresponding function arguments for the captured inputs.

  Returns:
    If the default graph is being used to define a function, the
    returned list of place holders are those used inside the function
    body corresponding those returned by get_extra_inputs(). Otherwise,
    returns an empty list.
  """
  g = ops.get_default_graph()
  if isinstance(g, _FuncGraph):
    return g.extra_args
  else:
    return []