def __init__(self,
               save_steps=None,
               save_secs=None,
               output_dir="",
               show_dataflow=True,
               show_memory=False):
    """Initializes a hook that takes periodic profiling snapshots.

    `options.run_metadata` argument of `tf.Session.Run` is used to collect
    metadata about execution. This hook sets the metadata and dumps it in Chrome
    Trace format.


    Args:
      save_steps: `int`, save profile traces every N steps. Exactly one of
          `save_secs` and `save_steps` should be set.
      save_secs: `int` or `float`, save profile traces every N seconds.
      output_dir: `string`, the directory to save the profile traces to.
          Defaults to the current directory.
      show_dataflow: `bool`, if True, add flow events to the trace connecting
          producers and consumers of tensors.
      show_memory: `bool`, if True, add object snapshot events to the trace
          showing the sizes and lifetimes of tensors.
    """
    self._output_file = os.path.join(output_dir, "timeline-{}.json")
    self._file_writer = SummaryWriterCache.get(output_dir)
    self._show_dataflow = show_dataflow
    self._show_memory = show_memory
    self._timer = SecondOrStepTimer(
        every_secs=save_secs, every_steps=save_steps)

  def __init__(self,
               save_steps=100,
               output_dir=None,
               summary_writer=None,
               scaffold=None,
               summary_op=None):
    """Initializes a `SummarySaver` monitor.

    Args:
      save_steps: `int`, save summaries every N steps. See `EveryN`.
      output_dir: `string`, the directory to save the summaries to. Only used
          if no `summary_writer` is supplied.
      summary_writer: `SummaryWriter`. If `None` and an `output_dir` was passed,
          one will be created accordingly.
      scaffold: `Scaffold` to get summary_op if it's not provided.
      summary_op: `Tensor` of type `string`. A serialized `Summary` protocol
          buffer, as output by TF summary methods like `scalar_summary` or
          `merge_all_summaries`.
    """
    # TODO(ipolosukhin): Implement every N seconds.
    self._summary_op = summary_op
    self._summary_writer = summary_writer
    if summary_writer is None and output_dir:
      self._summary_writer = SummaryWriterCache.get(output_dir)
    self._scaffold = scaffold
    self._save_steps = save_steps

  def __init__(self,
               checkpoint_dir,
               save_secs=None,
               save_steps=None,
               saver=None,
               checkpoint_basename="model.ckpt",
               scaffold=None):
    """Initialize CheckpointSaverHook monitor.

    Args:
      checkpoint_dir: `str`, base directory for the checkpoint files.
      save_secs: `int`, save every N secs.
      save_steps: `int`, save every N steps.
      saver: `Saver` object, used for saving.
      checkpoint_basename: `str`, base name for the checkpoint files.
      scaffold: `Scaffold`, use to get saver object.

    Raises:
      ValueError: One of `save_steps` or `save_secs` should be set.
      ValueError: Exactly one of saver or scaffold should be set.
    """
    logging.info("Create CheckpointSaverHook.")
    if ((saver is None and scaffold is None) or
        (saver is not None and scaffold is not None)):
      raise ValueError("Exactly one of saver or scaffold must be provided.")
    self._saver = saver
    self._checkpoint_dir = checkpoint_dir
    self._summary_writer = SummaryWriterCache.get(checkpoint_dir)
    self._save_path = os.path.join(checkpoint_dir, checkpoint_basename)
    self._scaffold = scaffold
    self._timer = _SecondOrStepTimer(every_secs=save_secs,
                                     every_steps=save_steps)

  def after_run(self, run_context, run_values):
    del run_context  # Unused by feature importance summary saver hook.

    # Read result tensors.
    global_step = run_values.results["global_step"]
    feature_names = run_values.results["feature_names"]
    feature_usage_counts = run_values.results["feature_usage_counts"]
    feature_gains = run_values.results["feature_gains"]

    # Ensure summaries are logged at desired frequency
    if (self._last_triggered_step is not None and
        global_step < self._last_triggered_step + self._every_n_steps):
      return

    # Validate tensors.
    if (len(feature_names) != len(feature_usage_counts) or
        len(feature_names) != len(feature_gains)):
      raise RuntimeError(
          "Feature names and importance measures have inconsistent lengths.")

    # Compute total usage.
    total_usage_count = 0.0
    for usage_count in feature_usage_counts:
      total_usage_count += usage_count
    usage_count_norm = 1.0 / total_usage_count if total_usage_count else 1.0

    # Compute total gain.
    total_gain = 0.0
    for gain in feature_gains:
      total_gain += gain
    gain_norm = 1.0 / total_gain if total_gain else 1.0

    # Output summary for each feature.
    self._last_triggered_step = global_step
    for (name, usage_count, gain) in zip(feature_names, feature_usage_counts,
                                         feature_gains):
      output_dir = os.path.join(self._model_dir, name.decode("utf-8"))
      summary_writer = SummaryWriterCache.get(output_dir)
      usage_count_summary = Summary(value=[
          Summary.Value(
              tag="feature_importance/usage_counts",
              simple_value=usage_count)
      ])
      usage_fraction_summary = Summary(value=[
          Summary.Value(
              tag="feature_importance/usage_fraction",
              simple_value=usage_count * usage_count_norm)
      ])
      summary_writer.add_summary(usage_count_summary, global_step)
      summary_writer.add_summary(usage_fraction_summary, global_step)
      gains_summary = Summary(
          value=[Summary.Value(
              tag="feature_importance/gains",
              simple_value=gain)])
      gains_fraction_summary = Summary(
          value=[Summary.Value(
              tag="feature_importance/gains_fraction",
              simple_value=gain * gain_norm)])
      summary_writer.add_summary(gains_summary, global_step)
      summary_writer.add_summary(gains_fraction_summary, global_step)

    def begin(self):
        # These calls only works because the SessionRunHook api guarantees this
        # will get called within a graph context containing our model graph.

        self.summary_writer = SummaryWriterCache.get(self.working_dir)
        self.weight_tensors = tf.trainable_variables()
        self.global_step = tf.train.get_or_create_global_step()

  def __init__(self,
               save_steps=100,
               save_secs=None,
               output_dir=None,
               summary_writer=None,
               scaffold=None,
               summary_op=None):
    """Initializes a `SummarySaver` monitor.

    Args:
      save_steps: `int`, save summaries every N steps. Exactly one of
          `save_secs` and `save_steps` should be set.
      save_secs: `int`, save summaries every N seconds.
      output_dir: `string`, the directory to save the summaries to. Only used
          if no `summary_writer` is supplied.
      summary_writer: `SummaryWriter`. If `None` and an `output_dir` was passed,
          one will be created accordingly.
      scaffold: `Scaffold` to get summary_op if it's not provided.
      summary_op: `Tensor` of type `string`. A serialized `Summary` protocol
          buffer, as output by TF summary methods like `scalar_summary` or
          `merge_all_summaries`.
    """
    self._summary_op = summary_op
    self._summary_writer = summary_writer
    if summary_writer is None and output_dir:
      self._summary_writer = SummaryWriterCache.get(output_dir)
    self._scaffold = scaffold
    self._timer = _SecondOrStepTimer(every_secs=save_secs,
                                     every_steps=save_steps)

 def begin(self):
   if self._summary_writer is None and self._output_dir:
     self._summary_writer = SummaryWriterCache.get(self._output_dir)
   self._next_step = None
   self._global_step_tensor = training_util.get_global_step()
   if self._global_step_tensor is None:
     raise RuntimeError(
         "Global step should be created to use SummarySaverHook.")

 def begin(self):
     if self._summary_writer is None and self._output_dir:
         self._summary_writer = SummaryWriterCache.get(self._output_dir)
     self._next_episode = None
     self._current_episode = None
     self._global_episode_tensor = get_global_episode()
     if self._global_episode_tensor is None:
         raise RuntimeError("Global episode should be created to use EpisodeSummarySaverHook.")

 def begin(self):
   self._summary_writer = SummaryWriterCache.get(self._checkpoint_dir)
   self._global_step_tensor = training_util._get_or_create_global_step_read()  # pylint: disable=protected-access
   if self._global_step_tensor is None:
     raise RuntimeError(
         "Global step should be created to use CheckpointSaverHook.")
   for l in self._listeners:
     l.begin()

 def begin(self):
   if self._summary_writer is None and self._output_dir:
     self._summary_writer = SummaryWriterCache.get(self._output_dir)
   self._next_step = None
   self._global_step_tensor = training_util._get_or_create_global_step_read()  # pylint: disable=protected-access
   if self._global_step_tensor is None:
     raise RuntimeError(
         "Global step should be created to use SummarySaverHook.")

  def __init__(self,
               every_n_steps=100,
               every_n_secs=None,
               output_dir=None,
               summary_writer=None):

    if (every_n_steps is None) == (every_n_secs is None):
      raise ValueError(
          "exactly one of every_n_steps and every_n_secs should be provided.")
    self._timer = _SecondOrStepTimer(every_steps=every_n_steps,
                                     every_secs=every_n_secs)

    self._summary_writer = summary_writer
    if summary_writer is None and output_dir:
      self._summary_writer = SummaryWriterCache.get(output_dir)

 def __init__(self, every_n_steps=100, output_dir=None, summary_writer=None):
   self._summary_tag = "global_step/sec"
   self._every_n_steps = every_n_steps
   self._summary_writer = summary_writer
   if summary_writer is None and output_dir:
     self._summary_writer = SummaryWriterCache.get(output_dir)