org.deeplearning4j.spark.api

Interface TrainingHook

All Superinterfaces:

java.io.Serializable

All Known Implementing Classes:

ParameterServerTrainingHook

public interface TrainingHook
extends java.io.Serializable

A hook for the workers when training. A pre update and post update method are specified for when certain information needs to be collected or there needs to be specific parameters or models sent to remote locations for visualization or other things.

Method Summary

Modifier and Type	Method and Description
void	postUpdate(org.nd4j.linalg.dataset.api.DataSet minibatch, Model model) A hook method for post update
void	postUpdate(org.nd4j.linalg.dataset.api.MultiDataSet minibatch, Model model) A hook method for post update
void	preUpdate(org.nd4j.linalg.dataset.api.DataSet minibatch, Model model) A hook method for pre update.
void	preUpdate(org.nd4j.linalg.dataset.api.MultiDataSet minibatch, Model model) A hook method for pre update.

Method Detail

preUpdate

void preUpdate(org.nd4j.linalg.dataset.api.DataSet minibatch,
               Model model)

A hook method for pre update.

Parameters:

minibatch - the inibatch that was used for the update

model - themodel that was update

postUpdate

void postUpdate(org.nd4j.linalg.dataset.api.DataSet minibatch,
                Model model)

A hook method for post update

Parameters:

minibatch - the minibatch that was usd for the update

model - the model that was updated

preUpdate

void preUpdate(org.nd4j.linalg.dataset.api.MultiDataSet minibatch,
               Model model)

A hook method for pre update.

Parameters:

minibatch - the inibatch that was used for the update

model - the model that was update

postUpdate

void postUpdate(org.nd4j.linalg.dataset.api.MultiDataSet minibatch,
                Model model)

A hook method for post update

Parameters:

minibatch - the minibatch that was usd for the update

model - the model that was updated