Package ai.djl.training.loss

Class SimpleCompositeLoss

java.lang.Object

ai.djl.training.evaluator.Evaluator

ai.djl.training.loss.Loss

ai.djl.training.loss.AbstractCompositeLoss

ai.djl.training.loss.SimpleCompositeLoss

public class SimpleCompositeLoss
extends AbstractCompositeLoss

SimpleCompositeLoss is an implementation of the Loss abstract class that can combine different Loss functions by adding the individual losses together.

For cases where the losses use only a single index of the labels and/or predictions, use the IndexLoss.

For an example of using this loss, see the captcha training example.

Field Summary

Fields inherited from class ai.djl.training.loss.AbstractCompositeLoss

components

Fields inherited from class ai.djl.training.evaluator.Evaluator

totalInstances

Constructor Summary

Constructors

Constructor	Description
SimpleCompositeLoss()	Creates a new empty instance of CompositeLoss that can combine the given Loss components.
SimpleCompositeLoss(String name)	Creates a new empty instance of CompositeLoss that can combine the given Loss components.

Method Summary

Modifier and Type	Method	Description
SimpleCompositeLoss	addLoss(Loss loss)	Adds a Loss that applies to all labels and predictions to this composite loss.
protected ai.djl.util.Pair<NDList,NDList>	inputForComponent(int componentIndex, NDList labels, NDList predictions)	Returns the inputs to computing the loss for a component loss.

Methods inherited from class ai.djl.training.loss.AbstractCompositeLoss

addAccumulator, evaluate, getAccumulator, getComponents, resetAccumulator, updateAccumulators

Methods inherited from class ai.djl.training.loss.Loss

elasticNetWeightedDecay, elasticNetWeightedDecay, elasticNetWeightedDecay, elasticNetWeightedDecay, hingeLoss, hingeLoss, hingeLoss, l1Loss, l1Loss, l1Loss, l1WeightedDecay, l1WeightedDecay, l1WeightedDecay, l2Loss, l2Loss, l2Loss, l2WeightedDecay, l2WeightedDecay, l2WeightedDecay, maskedSoftmaxCrossEntropyLoss, maskedSoftmaxCrossEntropyLoss, maskedSoftmaxCrossEntropyLoss, quantileL1Loss, quantileL1Loss, sigmoidBinaryCrossEntropyLoss, sigmoidBinaryCrossEntropyLoss, sigmoidBinaryCrossEntropyLoss, softmaxCrossEntropyLoss, softmaxCrossEntropyLoss, softmaxCrossEntropyLoss, updateAccumulator

Methods inherited from class ai.djl.training.evaluator.Evaluator

checkLabelShapes, checkLabelShapes, getName

Methods inherited from class java.lang.Object

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

Constructor Details

SimpleCompositeLoss

public SimpleCompositeLoss()

Creates a new empty instance of CompositeLoss that can combine the given Loss components.

SimpleCompositeLoss

public SimpleCompositeLoss(String name)

Creates a new empty instance of CompositeLoss that can combine the given Loss components.

Parameters:

name - the display name of the loss

Method Details

addLoss

public SimpleCompositeLoss addLoss(Loss loss)

Adds a Loss that applies to all labels and predictions to this composite loss.

Parameters:

loss - the loss to add

Returns:

this composite loss

inputForComponent

protected ai.djl.util.Pair<NDList,NDList> inputForComponent(int componentIndex,
 NDList labels,
 NDList predictions)

Returns the inputs to computing the loss for a component loss.

Specified by:

inputForComponent in class AbstractCompositeLoss

Parameters:

componentIndex - the index of the component loss

labels - the label input to the composite loss

predictions - the predictions input to the composite loss

Returns:

a pair of the (labels, predictions) inputs to the component loss