org.deeplearning4j.parallelism

Class EarlyStoppingParallelTrainer<T extends Model>

java.lang.Object

org.deeplearning4j.parallelism.EarlyStoppingParallelTrainer<T>

All Implemented Interfaces:

IEarlyStoppingTrainer<T>

public class EarlyStoppingParallelTrainer<T extends Model>
extends java.lang.Object
implements IEarlyStoppingTrainer<T>

Conduct parallel early stopping training with ParallelWrapper under the hood.
Can be used to train a MultiLayerNetwork or a ComputationGraph via early stopping.

Field Summary

Fields

Modifier and Type	Field and Description
protected EarlyStoppingConfiguration<T>	esConfig
protected T	model
protected IterationTerminationCondition	terminationReason

Constructor Summary

Constructors

Constructor and Description
EarlyStoppingParallelTrainer(EarlyStoppingConfiguration<T> earlyStoppingConfiguration, T model, org.nd4j.linalg.dataset.api.iterator.DataSetIterator train, org.nd4j.linalg.dataset.api.iterator.MultiDataSetIterator trainMulti, EarlyStoppingListener<T> listener, int workers, int prefetchBuffer, int averagingFrequency)
EarlyStoppingParallelTrainer(EarlyStoppingConfiguration<T> earlyStoppingConfiguration, T model, org.nd4j.linalg.dataset.api.iterator.DataSetIterator train, org.nd4j.linalg.dataset.api.iterator.MultiDataSetIterator trainMulti, EarlyStoppingListener<T> listener, int workers, int prefetchBuffer, int averagingFrequency, boolean reportScoreAfterAveraging, boolean useLegacyAveraging)
EarlyStoppingParallelTrainer(EarlyStoppingConfiguration<T> earlyStoppingConfiguration, T model, org.nd4j.linalg.dataset.api.iterator.DataSetIterator train, org.nd4j.linalg.dataset.api.iterator.MultiDataSetIterator trainMulti, int workers, int prefetchBuffer, int averagingFrequency)

Method Summary

Modifier and Type	Method and Description
EarlyStoppingResult<T>	fit() Conduct early stopping training
boolean	getTermination()
void	incrementIteration()
protected void	reset()
void	setLatestScore(double latestScore)
void	setListener(EarlyStoppingListener<T> listener) Set the early stopping listener
void	setTermination(boolean terminate)
protected void	setTerminationReason(IterationTerminationCondition terminationReason)

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Field Detail

model

protected T extends Model model

esConfig

protected final EarlyStoppingConfiguration<T extends Model> esConfig

terminationReason

protected volatile IterationTerminationCondition terminationReason

Constructor Detail

EarlyStoppingParallelTrainer

public EarlyStoppingParallelTrainer(EarlyStoppingConfiguration<T> earlyStoppingConfiguration,
                                    T model,
                                    org.nd4j.linalg.dataset.api.iterator.DataSetIterator train,
                                    org.nd4j.linalg.dataset.api.iterator.MultiDataSetIterator trainMulti,
                                    int workers,
                                    int prefetchBuffer,
                                    int averagingFrequency)

EarlyStoppingParallelTrainer

public EarlyStoppingParallelTrainer(EarlyStoppingConfiguration<T> earlyStoppingConfiguration,
                                    T model,
                                    org.nd4j.linalg.dataset.api.iterator.DataSetIterator train,
                                    org.nd4j.linalg.dataset.api.iterator.MultiDataSetIterator trainMulti,
                                    EarlyStoppingListener<T> listener,
                                    int workers,
                                    int prefetchBuffer,
                                    int averagingFrequency)

EarlyStoppingParallelTrainer

public EarlyStoppingParallelTrainer(EarlyStoppingConfiguration<T> earlyStoppingConfiguration,
                                    T model,
                                    org.nd4j.linalg.dataset.api.iterator.DataSetIterator train,
                                    org.nd4j.linalg.dataset.api.iterator.MultiDataSetIterator trainMulti,
                                    EarlyStoppingListener<T> listener,
                                    int workers,
                                    int prefetchBuffer,
                                    int averagingFrequency,
                                    boolean reportScoreAfterAveraging,
                                    boolean useLegacyAveraging)

Method Detail

setTerminationReason

protected void setTerminationReason(IterationTerminationCondition terminationReason)

fit

public EarlyStoppingResult<T> fit()

Description copied from interface: IEarlyStoppingTrainer

Conduct early stopping training

Specified by:

fit in interface IEarlyStoppingTrainer<T extends Model>

setLatestScore

public void setLatestScore(double latestScore)

incrementIteration

public void incrementIteration()

setTermination

public void setTermination(boolean terminate)

getTermination

public boolean getTermination()

setListener

public void setListener(EarlyStoppingListener<T> listener)

Description copied from interface: IEarlyStoppingTrainer

Set the early stopping listener

Specified by:

setListener in interface IEarlyStoppingTrainer<T extends Model>

reset

protected void reset()