org.deeplearning4j.parallelism.trainer

Class SymmetricTrainer

java.lang.Object

java.lang.Thread

org.deeplearning4j.parallelism.trainer.DefaultTrainer

org.deeplearning4j.parallelism.trainer.SymmetricTrainer

All Implemented Interfaces:

Runnable, CommunicativeTrainer, Trainer

public class SymmetricTrainer
extends DefaultTrainer
implements CommunicativeTrainer

This trainer implementation does parallel training via gradients broadcasts. After each iteration, gradients from this trainer will be propagated & applied to all other trainers

Author:

raver119@gmail.com

Nested Class Summary

Nested classes/interfaces inherited from class org.deeplearning4j.parallelism.trainer.DefaultTrainer

DefaultTrainer.DefaultTrainerBuilder

Nested classes/interfaces inherited from class java.lang.Thread

Thread.State, Thread.UncaughtExceptionHandler

Field Summary

Fields

Modifier and Type	Field and Description
protected org.deeplearning4j.optimize.solvers.accumulation.GradientsAccumulator	accumulator

Fields inherited from class org.deeplearning4j.parallelism.trainer.DefaultTrainer

averagingFrequency, isStopped, lastEtlTime, nullDataSet, nullMode, onRootModel, originalModel, parallelWrapper, queue, queueMDS, replicatedModel, running, shouldStop, shouldUpdate, threadId, thrownException, useMDS, uuid, workspaceMode

Fields inherited from class java.lang.Thread

MAX_PRIORITY, MIN_PRIORITY, NORM_PRIORITY

Constructor Summary

Constructors

Constructor and Description
SymmetricTrainer(org.deeplearning4j.nn.api.Model originalModel, int threadIdx, org.deeplearning4j.nn.conf.WorkspaceMode mode, ParallelWrapper wrapper, boolean useMDS)

Method Summary

Modifier and Type	Method and Description
boolean	averagingRequired() This method returns TRUE if this Trainer implementation assumes periodic aver
void	enqueueGradient(org.deeplearning4j.optimize.listeners.SharedGradient gradient) Deprecated.
protected void	postInit() This method does post-initialization configuration of Model.

Methods inherited from class org.deeplearning4j.parallelism.trainer.DefaultTrainer

cloneListener, configureListeners, feedDataSet, feedMultiDataSet, fit, fit, getModel, isRunning, run, setupIfNeccessary, shutdown, updateModel, waitTillRunning

Methods inherited from class java.lang.Thread

activeCount, checkAccess, clone, countStackFrames, currentThread, destroy, dumpStack, enumerate, getAllStackTraces, getContextClassLoader, getDefaultUncaughtExceptionHandler, getId, getName, getPriority, getStackTrace, getState, getThreadGroup, getUncaughtExceptionHandler, holdsLock, interrupt, interrupted, isAlive, isDaemon, isInterrupted, join, join, join, resume, setContextClassLoader, setDaemon, setDefaultUncaughtExceptionHandler, setName, setPriority, setUncaughtExceptionHandler, sleep, sleep, start, stop, stop, suspend, toString, yield

Methods inherited from class java.lang.Object

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

Methods inherited from interface org.deeplearning4j.parallelism.trainer.Trainer

feedDataSet, feedMultiDataSet, getModel, isRunning, setUncaughtExceptionHandler, shutdown, start, updateModel, waitTillRunning

Methods inherited from interface java.lang.Runnable

run

Field Detail

accumulator

protected org.deeplearning4j.optimize.solvers.accumulation.GradientsAccumulator accumulator

Constructor Detail

SymmetricTrainer

public SymmetricTrainer(@NonNull
                        org.deeplearning4j.nn.api.Model originalModel,
                        int threadIdx,
                        @NonNull
                        org.deeplearning4j.nn.conf.WorkspaceMode mode,
                        @NonNull
                        ParallelWrapper wrapper,
                        boolean useMDS)

Method Detail

enqueueGradient

@Deprecated
public void enqueueGradient(org.deeplearning4j.optimize.listeners.SharedGradient gradient)

Deprecated.

Specified by:

enqueueGradient in interface CommunicativeTrainer

averagingRequired

public boolean averagingRequired()

Description copied from interface: Trainer

This method returns TRUE if this Trainer implementation assumes periodic aver

Specified by:

averagingRequired in interface Trainer

Overrides:

averagingRequired in class DefaultTrainer

Returns:

postInit

protected void postInit()

Description copied from class: DefaultTrainer

This method does post-initialization configuration of Model. Good place to configure listeners and all such a things

Overrides:

postInit in class DefaultTrainer