Package org.nd4j.autodiff.listeners

Class BaseListener

java.lang.Object

org.nd4j.autodiff.listeners.BaseListener

All Implemented Interfaces:

Listener

Direct Known Subclasses:

ActivationGradientCheckListener, ArraySavingListener, BaseEvaluationListener, CheckpointListener, ControlflowListener, ExecDebuggingListener, OpBenchmarkListener, ProfilingListener, ScoreListener, UIListener

public abstract class BaseListener
extends Object
implements Listener

Constructor Summary

Constructors

Constructor	Description
BaseListener()

Method Summary

Modifier and Type	Method	Description
void	activationAvailable(SameDiff sd, At at, MultiDataSet batch, SameDiffOp op, String varName, INDArray activation)	Called when any activation becomes available.
ListenerResponse	epochEnd(SameDiff sd, At at, LossCurve lossCurve, long epochTimeMillis)	Called at the end of every epoch, when fitting from an iterator
void	epochStart(SameDiff sd, At at)	Called at the start of every epoch, when fitting from an iterator
void	iterationDone(SameDiff sd, At at, MultiDataSet dataSet, Loss loss)	Called at the end of every iteration, after all operations (including updating parameters) has been completed
void	iterationStart(SameDiff sd, At at, MultiDataSet data, long etlMs)	Called at the start of every iteration (minibatch), before any operations have been executed
void	operationEnd(SameDiff sd, Operation op)	Called at the end of an operation, e.g.
void	operationStart(SameDiff sd, Operation op)	Called at the start of an operation, e.g.
void	opExecution(SameDiff sd, At at, MultiDataSet batch, SameDiffOp op, OpContext opContext, INDArray[] outputs)	Called at the end of each operation execution
void	preOpExecution(SameDiff sd, At at, SameDiffOp op, OpContext opContext)	Called just before each operation is executed (native code called, etc) - after all inputs etc have been set
void	preUpdate(SameDiff sd, At at, Variable v, INDArray update)	Called just before each parameter is to be updated - i.e., just before each parameter is modified.
ListenerVariables	requiredVariables(SameDiff sd)	Required variables for this listener.
ListenerResponse	validationDone(SameDiff sd, At at, long validationTimeMillis)	Called after the end of every epoch, once validation evaluation is done, when training

Methods inherited from class java.lang.Object

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

Methods inherited from interface org.nd4j.autodiff.listeners.Listener

isActive

Constructor Detail

BaseListener

public BaseListener()

Method Detail

requiredVariables

public ListenerVariables requiredVariables(SameDiff sd)

Description copied from interface: Listener

Required variables for this listener.

Used to ensure these variables end up in the minimum required subgraph calculated by InferenceSession. Otherwise, if the variables weren't required by a loss variable, they would not be calculated.

Any variables in here are guaranteed to have Listener.activationAvailable(SameDiff, At, MultiDataSet, SameDiffOp, String, INDArray) called for them, regardless of whether they would normally be calculated or not.

Specified by:

requiredVariables in interface Listener

epochStart

public void epochStart(SameDiff sd,
                       At at)

Description copied from interface: Listener

Called at the start of every epoch, when fitting from an iterator

Specified by:

epochStart in interface Listener

Parameters:

sd - The SameDiff instance

at - Current iteration/epoch etc

epochEnd

public ListenerResponse epochEnd(SameDiff sd,
                                 At at,
                                 LossCurve lossCurve,
                                 long epochTimeMillis)

Description copied from interface: Listener

Called at the end of every epoch, when fitting from an iterator

Specified by:

epochEnd in interface Listener

Parameters:

sd - The SameDiff instance

at - Current iteration/epoch etc

lossCurve - The losses so far

epochTimeMillis - How long this epoch took

Returns:

ListenerResponse.STOP to stop training, CONTINUE or null to continue

validationDone

public ListenerResponse validationDone(SameDiff sd,
                                       At at,
                                       long validationTimeMillis)

Description copied from interface: Listener

Called after the end of every epoch, once validation evaluation is done, when training

Specified by:

validationDone in interface Listener

Parameters:

sd - The SameDiff instance

at - Current iteration/epoch etc

validationTimeMillis - How long validation took for this epoch

Returns:

ListenerResponse.STOP to stop training, CONTINUE or null to continue

iterationStart

public void iterationStart(SameDiff sd,
                           At at,
                           MultiDataSet data,
                           long etlMs)

Description copied from interface: Listener

Called at the start of every iteration (minibatch), before any operations have been executed

Specified by:

iterationStart in interface Listener

Parameters:

sd - The SameDiff instance

at - Current iteration/epoch etc

iterationDone

public void iterationDone(SameDiff sd,
                          At at,
                          MultiDataSet dataSet,
                          Loss loss)

Description copied from interface: Listener

Called at the end of every iteration, after all operations (including updating parameters) has been completed

Specified by:

iterationDone in interface Listener

Parameters:

sd - The SameDiff instance

at - Current iteration/epoch etc

dataSet - The current dataset (minibatch) used for training

loss - The loss value for the current minibatch. Will be null except for during training

operationStart

public void operationStart(SameDiff sd,
                           Operation op)

Description copied from interface: Listener

Called at the start of an operation, e.g. training or validation

Specified by:

operationStart in interface Listener

Parameters:

sd - The SameDiff instance

op - The operation being started

operationEnd

public void operationEnd(SameDiff sd,
                         Operation op)

Description copied from interface: Listener

Called at the end of an operation, e.g. training or validation

Specified by:

operationEnd in interface Listener

Parameters:

sd - The SameDiff instance

op - The operation being started

preOpExecution

public void preOpExecution(SameDiff sd,
                           At at,
                           SameDiffOp op,
                           OpContext opContext)

Description copied from interface: Listener

Called just before each operation is executed (native code called, etc) - after all inputs etc have been set

Specified by:

preOpExecution in interface Listener

Parameters:

sd - The SameDiff instance

at - Current iteration/epoch etc

op - Operation that has just been executed

opExecution

public void opExecution(SameDiff sd,
                        At at,
                        MultiDataSet batch,
                        SameDiffOp op,
                        OpContext opContext,
                        INDArray[] outputs)

Description copied from interface: Listener

Called at the end of each operation execution

Note: Outputs will most likely be freed later, use detach() if you need to save it.

Specified by:

opExecution in interface Listener

Parameters:

sd - The SameDiff instance

at - Current iteration/epoch etc

batch - The batch's input data. May be null if not called with a batch

op - Operation that has just been executed

outputs - The output arrays for the just-executed operation

activationAvailable

public void activationAvailable(SameDiff sd,
                                At at,
                                MultiDataSet batch,
                                SameDiffOp op,
                                String varName,
                                INDArray activation)

Description copied from interface: Listener

Called when any activation becomes available.

The activation will most likely be freed later, use dup() if you need to save it.

Note that this method will be called when any activation becomes available, not just ones from Listener.requiredVariables(SameDiff)
It is guaranteed to be called for variables from requiredVariables().

Note that the activations here overlap with Listener.opExecution(SameDiff, At, MultiDataSet, SameDiffOp, OpContext, INDArray[]) - both contain the same information/arrays

Specified by:

activationAvailable in interface Listener

Parameters:

sd - The SameDiff instance

at - Current iteration/epoch etc

batch - The batch's input data. May be null if not called with a batch

op - Operation that has just been executed

varName - The name of the variable

activation - The variable's activation

preUpdate

public void preUpdate(SameDiff sd,
                      At at,
                      Variable v,
                      INDArray update)

Description copied from interface: Listener

Called just before each parameter is to be updated - i.e., just before each parameter is modified.

Specified by:

preUpdate in interface Listener

Parameters:

sd - SameDiff instance

at - The current iteration/epoch etc

v - Variable about to be updated during backprop

update - The array representing the update (i.e., the gradient after applying learning rate, momentum, etc)