opennmt.utils.hooks module

Custom hooks.

class opennmt.utils.hooks.LogParametersCountHook[source]

Bases: tensorflow.python.training.session_run_hook.SessionRunHook

Simple hook that logs the number of trainable parameters.

begin()[source]

Called once before using the session.

When called, the default graph is the one that will be launched in the session. The hook can modify the graph by adding new operations to it. After the begin() call the graph will be finalized and the other callbacks can not modify the graph anymore. Second call of begin() on the same graph, should not change the graph.

opennmt.utils.hooks.add_counter(name, tensor)[source]

Registers a new counter.

Parameters:
  • name – The name of this counter.
  • tensor – The integer tf.Tensor to count.
Returns:

An op that increments the counter.

See also

opennmt.utils.misc.WordCounterHook() that fetches these counters to log their value in TensorBoard.

class opennmt.utils.hooks.CountersHook(every_n_steps=100, every_n_secs=None, output_dir=None, summary_writer=None, counters=None)[source]

Bases: tensorflow.python.training.session_run_hook.SessionRunHook

Hook that summarizes counters.

Implementation is mostly copied from StepCounterHook.

Deprecated: use LogWordsPerSecondHook instead.

begin()[source]

Called once before using the session.

When called, the default graph is the one that will be launched in the session. The hook can modify the graph by adding new operations to it. After the begin() call the graph will be finalized and the other callbacks can not modify the graph anymore. Second call of begin() on the same graph, should not change the graph.

before_run(run_context)[source]

Called before each call to run().

You can return from this call a SessionRunArgs object indicating ops or tensors to add to the upcoming run() call. These ops/tensors will be run together with the ops/tensors originally passed to the original run() call. The run args you return can also contain feeds to be added to the run() call.

The run_context argument is a SessionRunContext that provides information about the upcoming run() call: the originally requested op/tensors, the TensorFlow Session.

At this point graph is finalized and you can not add ops.

Parameters:run_context – A SessionRunContext object.
Returns:None or a SessionRunArgs object.
after_run(run_context, run_values)[source]

Called after each call to run().

The run_values argument contains results of requested ops/tensors by before_run().

The run_context argument is the same one send to before_run call. run_context.request_stop() can be called to stop the iteration.

If session.run() raises any exceptions then after_run() is not called.

Parameters:
  • run_context – A SessionRunContext object.
  • run_values – A SessionRunValues object.
class opennmt.utils.hooks.LogWordsPerSecondHook(num_words, every_n_steps=100, every_n_secs=None, output_dir=None, summary_writer=None)[source]

Bases: tensorflow.python.training.session_run_hook.SessionRunHook

Hook that logs the number of words processed per second.

Implementation is mostly copied from StepCounterHook.

begin()[source]

Called once before using the session.

When called, the default graph is the one that will be launched in the session. The hook can modify the graph by adding new operations to it. After the begin() call the graph will be finalized and the other callbacks can not modify the graph anymore. Second call of begin() on the same graph, should not change the graph.

after_create_session(session, coord)[source]

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.
Parameters:
  • session – A TensorFlow Session that has been created.
  • coord – A Coordinator object which keeps track of all threads.
before_run(run_context)[source]

Called before each call to run().

You can return from this call a SessionRunArgs object indicating ops or tensors to add to the upcoming run() call. These ops/tensors will be run together with the ops/tensors originally passed to the original run() call. The run args you return can also contain feeds to be added to the run() call.

The run_context argument is a SessionRunContext that provides information about the upcoming run() call: the originally requested op/tensors, the TensorFlow Session.

At this point graph is finalized and you can not add ops.

Parameters:run_context – A SessionRunContext object.
Returns:None or a SessionRunArgs object.
after_run(run_context, run_values)[source]

Called after each call to run().

The run_values argument contains results of requested ops/tensors by before_run().

The run_context argument is the same one send to before_run call. run_context.request_stop() can be called to stop the iteration.

If session.run() raises any exceptions then after_run() is not called.

Parameters:
  • run_context – A SessionRunContext object.
  • run_values – A SessionRunValues object.
class opennmt.utils.hooks.LogPredictionTimeHook[source]

Bases: tensorflow.python.training.session_run_hook.SessionRunHook

Hooks that gathers and logs prediction times.

begin()[source]

Called once before using the session.

When called, the default graph is the one that will be launched in the session. The hook can modify the graph by adding new operations to it. After the begin() call the graph will be finalized and the other callbacks can not modify the graph anymore. Second call of begin() on the same graph, should not change the graph.

before_run(run_context)[source]

Called before each call to run().

You can return from this call a SessionRunArgs object indicating ops or tensors to add to the upcoming run() call. These ops/tensors will be run together with the ops/tensors originally passed to the original run() call. The run args you return can also contain feeds to be added to the run() call.

The run_context argument is a SessionRunContext that provides information about the upcoming run() call: the originally requested op/tensors, the TensorFlow Session.

At this point graph is finalized and you can not add ops.

Parameters:run_context – A SessionRunContext object.
Returns:None or a SessionRunArgs object.
after_run(run_context, run_values)[source]

Called after each call to run().

The run_values argument contains results of requested ops/tensors by before_run().

The run_context argument is the same one send to before_run call. run_context.request_stop() can be called to stop the iteration.

If session.run() raises any exceptions then after_run() is not called.

Parameters:
  • run_context – A SessionRunContext object.
  • run_values – A SessionRunValues object.
end(session)[source]

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.

Parameters:session – A TensorFlow Session that will be soon closed.
class opennmt.utils.hooks.SaveEvaluationPredictionHook(model, output_file, post_evaluation_fn=None, predictions=None)[source]

Bases: tensorflow.python.training.session_run_hook.SessionRunHook

Hook that saves the evaluation predictions.

__init__(model, output_file, post_evaluation_fn=None, predictions=None)[source]

Initializes this hook.

Parameters:
  • model – The model for which to save the evaluation predictions.
  • output_file – The output filename which will be suffixed by the current training step.
  • post_evaluation_fn – (optional) A callable that takes as argument the current step and the file with the saved predictions.
  • predictions – The predictions to save.
begin()[source]

Called once before using the session.

When called, the default graph is the one that will be launched in the session. The hook can modify the graph by adding new operations to it. After the begin() call the graph will be finalized and the other callbacks can not modify the graph anymore. Second call of begin() on the same graph, should not change the graph.

before_run(run_context)[source]

Called before each call to run().

You can return from this call a SessionRunArgs object indicating ops or tensors to add to the upcoming run() call. These ops/tensors will be run together with the ops/tensors originally passed to the original run() call. The run args you return can also contain feeds to be added to the run() call.

The run_context argument is a SessionRunContext that provides information about the upcoming run() call: the originally requested op/tensors, the TensorFlow Session.

At this point graph is finalized and you can not add ops.

Parameters:run_context – A SessionRunContext object.
Returns:None or a SessionRunArgs object.
after_run(run_context, run_values)[source]

Called after each call to run().

The run_values argument contains results of requested ops/tensors by before_run().

The run_context argument is the same one send to before_run call. run_context.request_stop() can be called to stop the iteration.

If session.run() raises any exceptions then after_run() is not called.

Parameters:
  • run_context – A SessionRunContext object.
  • run_values – A SessionRunValues object.
end(session)[source]

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.

Parameters:session – A TensorFlow Session that will be soon closed.
class opennmt.utils.hooks.LoadWeightsFromCheckpointHook(checkpoint_path)[source]

Bases: tensorflow.python.training.session_run_hook.SessionRunHook

“Hook that loads model variables from checkpoint before starting the training.

begin()[source]

Called once before using the session.

When called, the default graph is the one that will be launched in the session. The hook can modify the graph by adding new operations to it. After the begin() call the graph will be finalized and the other callbacks can not modify the graph anymore. Second call of begin() on the same graph, should not change the graph.

after_create_session(session, coord)[source]

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.
Parameters:
  • session – A TensorFlow Session that has been created.
  • coord – A Coordinator object which keeps track of all threads.
class opennmt.utils.hooks.VariablesInitializerHook(variables)[source]

Bases: tensorflow.python.training.session_run_hook.SessionRunHook

Hook that initializes some variables in the current session. This is useful when using internal variables (e.g. for value accumulation) that are not saved in the checkpoints.

__init__(variables)[source]

Initializes this hook.

Parameters:variables – A list of variables to initialize.
begin()[source]

Called once before using the session.

When called, the default graph is the one that will be launched in the session. The hook can modify the graph by adding new operations to it. After the begin() call the graph will be finalized and the other callbacks can not modify the graph anymore. Second call of begin() on the same graph, should not change the graph.

after_create_session(session, coord)[source]

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