Python tensorflow.python.platform.tf_logging.warning() Examples

def _clip_sparse(self, grad, var):
    assert isinstance(grad, ops.IndexedSlices)
    clip_dims = self._vars_to_clip_dims[var]
    if 0 in clip_dims:
      logging.warning("Clipping norm across dims %s for %s is inefficient "
                      "when including sparse dimension 0.", clip_dims,
                      var.op.name)
      return self._clip_dense(var)

    with ops.colocate_with(var):
      var_subset = array_ops.gather(var, grad.indices)
    with self._maybe_colocate_with(var):
      normalized_var_subset = clip_ops.clip_by_norm(
          var_subset, self._max_norm, clip_dims)
      delta = ops.IndexedSlices(
          var_subset - normalized_var_subset, grad.indices, grad.dense_shape)
    with ops.colocate_with(var):
      return var.scatter_sub(delta, use_locking=self._use_locking) 

def _extract_metric_update_ops(self, eval_dict):
    """Separate update operations from metric value operations."""
    update_ops = []
    value_ops = {}
    for name, metric_ops in six.iteritems(eval_dict):
      if isinstance(metric_ops, (list, tuple)):
        if len(metric_ops) == 2:
          value_ops[name] = metric_ops[0]
          update_ops.append(metric_ops[1])
        else:
          logging.warning(
              'Ignoring metric {}. It returned a list|tuple with len {}, '
              'expected 2'.format(name, len(metric_ops)))
          value_ops[name] = metric_ops
      else:
        value_ops[name] = metric_ops

    if update_ops:
      update_ops = control_flow_ops.group(*update_ops)
    else:
      update_ops = None

    return update_ops, value_ops 

def prepend_name_scope(name, import_scope):
  """Prepends name scope to a name.

  Args:
    name: A `string` name.
    import_scope: Optional `string`. Name scope to add.

  Returns:
    Name with name scope added, or the original name if import_scope
    is None.
  """
  if import_scope:
    try:
      str_to_replace = r"([\^]|loc:@|^)(.*)"
      return re.sub(str_to_replace, r"\1" + import_scope + r"/\2",
                    compat.as_str(name))
    except TypeError as e:
      # If the name is not of a type we can process, simply return it.
      logging.warning(e)
      return name
  else:
    return name


# pylint: disable=g-doc-return-or-yield 

def strip_name_scope(name, export_scope):
  """Removes name scope from a name.

  Args:
    name: A `string` name.
    export_scope: Optional `string`. Name scope to remove.

  Returns:
    Name with name scope removed, or the original name if export_scope
    is None.
  """
  if export_scope:
    try:
      # Strips export_scope/, export_scope///,
      # ^export_scope/, loc:@export_scope/.
      str_to_replace = r"([\^]|loc:@|^)" + export_scope + r"[\/]+(.*)"
      return re.sub(str_to_replace, r"\1\2", compat.as_str(name), count=1)
    except TypeError as e:
      # If the name is not of a type we can process, simply return it.
      logging.warning(e)
      return name
  else:
    return name 

def _validate_input_pipeline(self):
    """Validates the input pipeline.

    Perform some sanity checks to log user friendly information. We should
    error out to give users better error message. But, if
    _WRAP_INPUT_FN_INTO_WHILE_LOOP is False (legacy behavior), we cannot break
    user code, so, log a warning.

    Raises:
      RuntimeError: If the validation failed.
    """
    if ops.get_default_graph().get_collection(ops.GraphKeys.QUEUE_RUNNERS):
      err_msg = ('Input pipeline contains one or more QueueRunners. '
                 'It could be slow and not scalable. Please consider '
                 'converting your input pipeline to use `tf.data` instead (see '
                 'https://www.tensorflow.org/guide/datasets for '
                 'instructions.')
      if _WRAP_INPUT_FN_INTO_WHILE_LOOP:
        raise RuntimeError(err_msg)
      else:
        logging.warn(err_msg) 

def _MakeShape(v, arg_name):
  """Convert v into a TensorShapeProto."""
  # Args:
  #   v: A TensorShapeProto, a list of ints, or a tensor_shape.TensorShape.
  #   arg_name: String, for error messages.

  # Returns:
  #   A TensorShapeProto.
  if isinstance(v, tensor_shape_pb2.TensorShapeProto):
    for d in v.dim:
      if d.name:
        logging.warning("Warning: TensorShapeProto with a named dimension: %s",
                        str(v))
        break
    return v
  return tensor_shape.as_shape(v).as_proto() 

def _load_library(name, op_list=None):
  """Loads a .so file containing the specified operators.

  Args:
    name: The name of the .so file to load.
    op_list: A list of names of operators that the library should have. If None
        then the .so file's contents will not be verified.

  Raises:
    NameError if one of the required ops is missing.
  """
  try:
    filename = resource_loader.get_path_to_datafile(name)
    library = load_library.load_op_library(filename)
    for expected_op in (op_list or []):
      for lib_op in library.OP_LIST.op:
        if lib_op.name == expected_op:
          break
      else:
        raise NameError('Could not find operator %s in dynamic library %s' %
                        (expected_op, name))
  except errors.NotFoundError:
    logging.warning('%s file could not be loaded.', name) 

def create_cpu_hostcall(host_calls):
    """Runs on the host_call on CPU instead of TPU when use_tpu=False."""

    _OutfeedHostCall.validate(host_calls)
    ret = {}
    for name, host_call in host_calls.items():
      host_fn, tensors = host_call
      if isinstance(tensors, (tuple, list)):
        ret[name] = host_fn(*tensors)
      else:
        # Must be dict.
        try:
          ret[name] = host_fn(**tensors)
        except TypeError as e:
          logging.warning(
              'Exception while calling %s: %s. It is likely the tensors '
              '(%s[1]) do not match the '
              'function\'s arguments', name, e, name)
          raise e
    return ret 

def _serve_runs(self, unused_query_params):
    """Return a JSON object about runs and tags.

    Returns a mapping from runs to tagType to list of tags for that run.

    Returns:
      {runName: {images: [tag1, tag2, tag3],
                 audio: [tag4, tag5, tag6],
                 scalars: [tagA, tagB, tagC],
                 histograms: [tagX, tagY, tagZ],
                 firstEventTimestamp: 123456.789}}
    """
    runs = self._multiplexer.Runs()
    for run_name, run_data in runs.items():
      try:
        run_data['firstEventTimestamp'] = self._multiplexer.FirstEventTimestamp(
            run_name)
      except ValueError:
        logging.warning('Unable to get first event timestamp for run %s',
                        run_name)
        run_data['firstEventTimestamp'] = None
    self.respond(runs, 'application/json') 

def _clip_sparse(self, grad, var):
    assert isinstance(grad, ops.IndexedSlices)
    clip_dims = self._vars_to_clip_dims[var]
    if 0 in clip_dims:
      logging.warning("Clipping norm across dims %s for %s is inefficient "
                      "when including sparse dimension 0.", clip_dims,
                      var.op.name)
      return self._clip_dense(var)

    with ops.colocate_with(var):
      var_subset = array_ops.gather(var, grad.indices)
    with self._maybe_colocate_with(var):
      normalized_var_subset = clip_ops.clip_by_norm(
          var_subset, self._max_norm, clip_dims)
      delta = ops.IndexedSlices(
          var_subset - normalized_var_subset, grad.indices, grad.dense_shape)
    with ops.colocate_with(var):
      return var.scatter_sub(delta, use_locking=self._use_locking) 

def _MakeShape(v, arg_name):
  """Convert v into a TensorShapeProto."""
  # Args:
  #   v: A TensorShapeProto, a list of ints, or a tensor_shape.TensorShape.
  #   arg_name: String, for error messages.

  # Returns:
  #   A TensorShapeProto.
  if isinstance(v, tensor_shape_pb2.TensorShapeProto):
    for d in v.dim:
      if d.name:
        logging.warning("Warning: TensorShapeProto with a named dimension: %s",
                        str(v))
        break
    return v
  return tensor_shape.as_shape(v).as_proto() 

def _MakeShape(v, arg_name):
  """Convert v into a TensorShapeProto."""
  # Args:
  #   v: A TensorShapeProto, a list of ints, or a tensor_shape.TensorShape.
  #   arg_name: String, for error messages.

  # Returns:
  #   A TensorShapeProto.
  if isinstance(v, tensor_shape_pb2.TensorShapeProto):
    for d in v.dim:
      if d.name:
        logging.warning("Warning: TensorShapeProto with a named dimension: %s",
                        str(v))
        break
    return v
  return tensor_shape.as_shape(v).as_proto() 

def _verify_model_fn_args(model_fn, params):
  """Verifies model fn arguments."""
  args = set(_model_fn_args(model_fn))
  if 'features' not in args:
    raise ValueError('model_fn (%s) must include features argument.' % model_fn)
  if 'labels' not in args:
    raise ValueError('model_fn (%s) must include labels argument.' % model_fn)
  if params is not None and 'params' not in args:
    raise ValueError('model_fn (%s) does not include params argument, '
                     'but params (%s) is passed to Estimator.' % (model_fn,
                                                                  params))
  if params is None and 'params' in args:
    logging.warning('Estimator\'s model_fn (%s) includes params '
                    'argument, but params are not passed to Estimator.',
                    model_fn)
  if tf_inspect.ismethod(model_fn):
    if 'self' in args:
      args.remove('self')
  non_valid_args = list(args - _VALID_MODEL_FN_ARGS)
  if non_valid_args:
    raise ValueError('model_fn (%s) has following not expected args: %s' %
                     (model_fn, non_valid_args)) 

def _extract_metric_update_ops(self, eval_dict):
    """Separate update operations from metric value operations."""
    update_ops = []
    value_ops = {}
    for name, metric_ops in six.iteritems(eval_dict):
      if isinstance(metric_ops, (list, tuple)):
        if len(metric_ops) == 2:
          value_ops[name] = metric_ops[0]
          update_ops.append(metric_ops[1])
        else:
          logging.warning(
              'Ignoring metric {}. It returned a list|tuple with len {}, '
              'expected 2'.format(name, len(metric_ops)))
          value_ops[name] = metric_ops
      else:
        value_ops[name] = metric_ops

    if update_ops:
      update_ops = control_flow_ops.group(*update_ops)
    else:
      update_ops = None

    return update_ops, value_ops 

def _serve_runs(self, unused_query_params):
    """Return a JSON object about runs and tags.

    Returns a mapping from runs to tagType to list of tags for that run.

    Returns:
      {runName: {images: [tag1, tag2, tag3],
                 audio: [tag4, tag5, tag6],
                 scalars: [tagA, tagB, tagC],
                 histograms: [tagX, tagY, tagZ],
                 firstEventTimestamp: 123456.789}}
    """
    runs = self._multiplexer.Runs()
    for run_name, run_data in runs.items():
      try:
        run_data['firstEventTimestamp'] = self._multiplexer.FirstEventTimestamp(
            run_name)
      except ValueError:
        logging.warning('Unable to get first event timestamp for run %s',
                        run_name)
        run_data['firstEventTimestamp'] = None
    self.respond(runs, 'application/json') 

def after_run(self, run_context, run_values):
    if np.isnan(run_values.results):
      failure_message = "Model diverged with loss = NaN."
      if self._fail_on_nan_loss:
        logging.error(failure_message)
        raise NanLossDuringTrainingError
      else:
        logging.warning(failure_message)
        # We don't raise an error but we request stop without an exception.
        run_context.request_stop() 

def _serve_static_file(self, path):
    """Serves the static file located at the given path.

    Args:
      path: The path of the static file, relative to the tensorboard/ directory.
    """
    # Strip off the leading forward slash.
    orig_path = path.lstrip('/')
    if not self._path_is_safe(orig_path):
      logging.warning('path not safe: %s', orig_path)
      self.respond('Naughty naughty!', 'text/plain', 400)
      return
    # Resource loader wants a path relative to //WORKSPACE/tensorflow.
    path = os.path.join('tensorboard', orig_path)
    # Open the file and read it.
    try:
      contents = resource_loader.load_resource(path)
    except IOError:
      # For compatibility with latest version of Bazel, we renamed bower
      # packages to use '_' rather than '-' in their package name.
      # This means that the directory structure is changed too.
      # So that all our recursive imports work, we need to modify incoming
      # requests to map onto the new directory structure.
      path = orig_path
      components = path.split('/')
      components[0] = components[0].replace('-', '_')
      path = ('/').join(components)
      # Bazel keeps all the external dependencies in //WORKSPACE/external.
      # and resource loader wants a path relative to //WORKSPACE/tensorflow/.
      path = os.path.join('../external', path)
      try:
        contents = resource_loader.load_resource(path)
      except IOError:
        logging.info('path %s not found, sending 404', path)
        self.respond('Not found', 'text/plain', 404)
        return
    mimetype, content_encoding = mimetypes.guess_type(path)
    mimetype = mimetype or 'application/octet-stream'
    self.respond(contents, mimetype, expires=3600,
                 content_encoding=content_encoding) 

def _default_global_step_tensor(self):
    """Returns the global_step from the default graph.

    Returns:
      The global step `Tensor` or `None`.
    """
    try:
      gs = ops.get_default_graph().get_tensor_by_name("global_step:0")
      if gs.dtype.base_dtype in [dtypes.int32, dtypes.int64]:
        return gs
      else:
        logging.warning("Found 'global_step' is not an int type: %s", gs.dtype)
        return None
    except KeyError:
      return None 

def _setup_data_handlers(self):
    self.data_handlers = {
        DATA_PREFIX + LOGDIR_ROUTE: self._serve_logdir,
        DATA_PREFIX + SCALARS_ROUTE: self._serve_scalars,
        DATA_PREFIX + GRAPH_ROUTE: self._serve_graph,
        DATA_PREFIX + RUN_METADATA_ROUTE: self._serve_run_metadata,
        DATA_PREFIX + HISTOGRAMS_ROUTE: self._serve_histograms,
        DATA_PREFIX + COMPRESSED_HISTOGRAMS_ROUTE:
            self._serve_compressed_histograms,
        DATA_PREFIX + IMAGES_ROUTE: self._serve_images,
        DATA_PREFIX + INDIVIDUAL_IMAGE_ROUTE: self._serve_image,
        DATA_PREFIX + AUDIO_ROUTE: self._serve_audio,
        DATA_PREFIX + INDIVIDUAL_AUDIO_ROUTE: self._serve_individual_audio,
        DATA_PREFIX + RUNS_ROUTE: self._serve_runs,
        '/app.js': self._serve_js
    }

    # Serve the routes from the registered plugins using their name as the route
    # prefix. For example if plugin z has two routes /a and /b, they will be
    # served as /data/plugin/z/a and /data/plugin/z/b.
    for name in REGISTERED_PLUGINS:
      try:
        plugin = REGISTERED_PLUGINS[name]
        plugin_handlers = plugin.get_plugin_handlers(
            self._multiplexer.RunPaths(), self._logdir)
      except Exception as e:  # pylint: disable=broad-except
        logging.warning('Plugin %s failed. Exception: %s', name, str(e))
        continue
      for route, handler in plugin_handlers.items():
        path = DATA_PREFIX + PLUGIN_PREFIX + '/' + name + route
        self.data_handlers[path] = functools.partial(handler, self) 

def VARIABLES(cls):  # pylint: disable=no-self-argument
    logging.warning("VARIABLES collection name is deprecated, "
                    "please use GLOBAL_VARIABLES instead; "
                    "VARIABLES will be removed after 2017-03-02.")
    return cls.GLOBAL_VARIABLES 

def AddRun(self, path, name=None):
    """Add a run to the multiplexer.

    If the name is not specified, it is the same as the path.

    If a run by that name exists, and we are already watching the right path,
      do nothing. If we are watching a different path, replace the event
      accumulator.

    If `Reload` has been called, it will `Reload` the newly created
    accumulators.

    Args:
      path: Path to the event files (or event directory) for given run.
      name: Name of the run to add. If not provided, is set to path.

    Returns:
      The `EventMultiplexer`.
    """
    if name is None or name is '':
      name = path
    accumulator = None
    with self._accumulators_mutex:
      if name not in self._accumulators or self._paths[name] != path:
        if name in self._paths and self._paths[name] != path:
          # TODO(danmane) - Make it impossible to overwrite an old path with
          # a new path (just give the new path a distinct name)
          logging.warning('Conflict for name %s: old path %s, new path %s',
                          name, self._paths[name], path)
        logging.info('Constructing EventAccumulator for %s', path)
        accumulator = event_accumulator.EventAccumulator(
            path,
            size_guidance=self._size_guidance,
            purge_orphaned_data=self.purge_orphaned_data)
        self._accumulators[name] = accumulator
        self._paths[name] = path
    if accumulator:
      if self._reload_called:
        accumulator.Reload()
    return self 

def _default_global_step_tensor(self):
    """Returns the global_step from the default graph.

    Returns:
      The global step `Tensor` or `None`.
    """
    try:
      gs = ops.get_default_graph().get_tensor_by_name("global_step:0")
      if gs.dtype.base_dtype in [dtypes.int32, dtypes.int64]:
        return gs
      else:
        logging.warning("Found 'global_step' is not an int type: %s", gs.dtype)
        return None
    except KeyError:
      return None 

def experimental(func):
  """Decorator for marking functions or methods experimental.

  This decorator logs an experimental warning whenever the decorated function is
  called. It has the following format:

    <function> (from <module>) is experimental and may change or be removed at
    any time, and without warning.

  <function> will include the class name if it is a method.

  It also edits the docstring of the function: ' (experimental)' is appended
  to the first line of the docstring and a notice is prepended to the rest of
  the docstring.

  Args:
    func: A function or method to mark experimental.

  Returns:
    Decorated function or method.
  """
  decorator_utils.validate_callable(func, 'experimental')
  @functools.wraps(func)
  def new_func(*args, **kwargs):
    logging.warning(
        '%s (from %s) is experimental and may change or be removed at '
        'any time, and without warning.',
        decorator_utils.get_qualified_name(func), func.__module__)
    return func(*args, **kwargs)
  new_func.__doc__ = _add_experimental_function_notice_to_docstring(
      func.__doc__)
  return new_func 

def every_n_step_end(self, step, outputs):
    super(NanLoss, self).every_n_step_end(step, outputs)
    if np.isnan(_extract_output(outputs, self._loss_tensor)):
      failure_message = "Model diverged with loss = NaN."
      if self._fail_on_nan_loss:
        logging.error(failure_message)
        raise NanLossDuringTrainingError
      else:
        logging.warning(failure_message)
        # We don't raise an error but we return "should stop" so we stop, but
        # without an exception.
        return True 

def _get_reader_for_run(self, run):
    if run in self.readers:
      return self.readers[run]

    config = self._configs[run]
    reader = None
    if config.model_checkpoint_path:
      try:
        reader = NewCheckpointReader(config.model_checkpoint_path)
      except Exception:  # pylint: disable=broad-except
        logging.warning('Failed reading %s', config.model_checkpoint_path)
    self.readers[run] = reader
    return reader 

def _serve_static_file(self, path):
    """Serves the static file located at the given path.

    Args:
      path: The path of the static file, relative to the tensorboard/ directory.
    """
    # Strip off the leading forward slash.
    orig_path = path.lstrip('/')
    if not self._path_is_safe(orig_path):
      logging.warning('path not safe: %s', orig_path)
      self.respond('Naughty naughty!', 'text/plain', 400)
      return
    # Resource loader wants a path relative to //WORKSPACE/tensorflow.
    path = os.path.join('tensorboard', orig_path)
    # Open the file and read it.
    try:
      contents = resource_loader.load_resource(path)
    except IOError:
      # For compatibility with latest version of Bazel, we renamed bower
      # packages to use '_' rather than '-' in their package name.
      # This means that the directory structure is changed too.
      # So that all our recursive imports work, we need to modify incoming
      # requests to map onto the new directory structure.
      path = orig_path
      components = path.split('/')
      components[0] = components[0].replace('-', '_')
      path = ('/').join(components)
      # Bazel keeps all the external dependencies in //WORKSPACE/external.
      # and resource loader wants a path relative to //WORKSPACE/tensorflow/.
      path = os.path.join('../external', path)
      try:
        contents = resource_loader.load_resource(path)
      except IOError:
        logging.info('path %s not found, sending 404', path)
        self.respond('Not found', 'text/plain', 404)
        return
    mimetype, content_encoding = mimetypes.guess_type(path)
    mimetype = mimetype or 'application/octet-stream'
    self.respond(contents, mimetype, expires=3600,
                 content_encoding=content_encoding) 

def _setup_data_handlers(self):
    self.data_handlers = {
        DATA_PREFIX + LOGDIR_ROUTE: self._serve_logdir,
        DATA_PREFIX + SCALARS_ROUTE: self._serve_scalars,
        DATA_PREFIX + GRAPH_ROUTE: self._serve_graph,
        DATA_PREFIX + RUN_METADATA_ROUTE: self._serve_run_metadata,
        DATA_PREFIX + HISTOGRAMS_ROUTE: self._serve_histograms,
        DATA_PREFIX + COMPRESSED_HISTOGRAMS_ROUTE:
            self._serve_compressed_histograms,
        DATA_PREFIX + IMAGES_ROUTE: self._serve_images,
        DATA_PREFIX + INDIVIDUAL_IMAGE_ROUTE: self._serve_image,
        DATA_PREFIX + AUDIO_ROUTE: self._serve_audio,
        DATA_PREFIX + INDIVIDUAL_AUDIO_ROUTE: self._serve_individual_audio,
        DATA_PREFIX + RUNS_ROUTE: self._serve_runs,
        '/app.js': self._serve_js
    }

    # Serve the routes from the registered plugins using their name as the route
    # prefix. For example if plugin z has two routes /a and /b, they will be
    # served as /data/plugin/z/a and /data/plugin/z/b.
    for name in self._registered_plugins:
      try:
        plugin = self._registered_plugins[name]
        plugin_handlers = plugin.get_plugin_handlers(
            self._multiplexer.RunPaths(), self._logdir)
      except Exception as e:  # pylint: disable=broad-except
        logging.warning('Plugin %s failed. Exception: %s', name, str(e))
        continue
      for route, handler in plugin_handlers.items():
        path = DATA_PREFIX + PLUGIN_PREFIX + '/' + name + route
        self.data_handlers[path] = functools.partial(handler, self) 

def VARIABLES(cls):  # pylint: disable=no-self-argument
    logging.warning("VARIABLES collection name is deprecated, "
                    "please use GLOBAL_VARIABLES instead; "
                    "VARIABLES will be removed after 2017-03-02.")
    return cls.GLOBAL_VARIABLES 

def AddRun(self, path, name=None):
    """Add a run to the multiplexer.

    If the name is not specified, it is the same as the path.

    If a run by that name exists, and we are already watching the right path,
      do nothing. If we are watching a different path, replace the event
      accumulator.

    If `Reload` has been called, it will `Reload` the newly created
    accumulators.

    Args:
      path: Path to the event files (or event directory) for given run.
      name: Name of the run to add. If not provided, is set to path.

    Returns:
      The `EventMultiplexer`.
    """
    if name is None or name is '':
      name = path
    accumulator = None
    with self._accumulators_mutex:
      if name not in self._accumulators or self._paths[name] != path:
        if name in self._paths and self._paths[name] != path:
          # TODO(danmane) - Make it impossible to overwrite an old path with
          # a new path (just give the new path a distinct name)
          logging.warning('Conflict for name %s: old path %s, new path %s',
                          name, self._paths[name], path)
        logging.info('Constructing EventAccumulator for %s', path)
        accumulator = event_accumulator.EventAccumulator(
            path,
            size_guidance=self._size_guidance,
            purge_orphaned_data=self.purge_orphaned_data)
        self._accumulators[name] = accumulator
        self._paths[name] = path
    if accumulator:
      if self._reload_called:
        accumulator.Reload()
    return self 

def _default_global_step_tensor(self):
    """Returns the global_step from the default graph.

    Returns:
      The global step `Tensor` or `None`.
    """
    try:
      gs = ops.get_default_graph().get_tensor_by_name("global_step:0")
      if gs.dtype.base_dtype in [dtypes.int32, dtypes.int64]:
        return gs
      else:
        logging.warning("Found 'global_step' is not an int type: %s", gs.dtype)
        return None
    except KeyError:
      return None