Join us at TensorFlow World, Oct 28-31. Use code TF20 for 20% off select passes. Register now

tf.train.ProfilerHook

View source on GitHub

Class ProfilerHook

Captures CPU/GPU profiling information every N steps or seconds.

Inherits From: SessionRunHook

Aliases:

  • Class tf.compat.v1.estimator.ProfilerHook
  • Class tf.compat.v1.train.ProfilerHook
  • Class tf.compat.v2.estimator.ProfilerHook
  • Class tf.estimator.ProfilerHook

This produces files called "timeline-.json", which are in Chrome Trace format.

For more information see: https://github.com/catapult-project/catapult/blob/master/tracing/README.md

__init__

View source

__init__(
    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.

Methods

after_create_session

View source

after_create_session(
    session,
    coord
)

Called when new TensorFlow session is created.

This is called to signal the hooks that a new session has been created. This has two essential differences with the situation in which begin is called:

  • When this is called, the graph is finalized and ops can no longer be added to the graph.
  • This method will also be called as a result of recovering a wrapped session, not only at the beginning of the overall session.

Args:

  • session: A TensorFlow Session that has been created.
  • coord: A Coordinator object which keeps track of all threads.

after_run

View source

after_run(
    run_context,
    run_values
)

before_run

View source

before_run(run_context)

begin

View source

begin()

end

View source

end(session)

Called at the end of session.

The session argument can be used in case the hook wants to run final ops, such as saving a last checkpoint.

If session.run() raises exception other than OutOfRangeError or StopIteration then end() is not called. Note the difference between end() and after_run() behavior when session.run() raises OutOfRangeError or StopIteration. In that case end() is called but after_run() is not called.

Args:

  • session: A TensorFlow Session that will be soon closed.