ai.djl.mxnet.engine

Class MxGradientCollector

java.lang.Object

ai.djl.mxnet.engine.MxGradientCollector

All Implemented Interfaces:

ai.djl.training.GradientCollector, java.lang.AutoCloseable

public class MxGradientCollector
extends java.lang.Object
implements ai.djl.training.GradientCollector

MxGradientCollector is the MXNet implementation of GradientCollector.

Constructor Summary

Constructors

Constructor and Description
MxGradientCollector() Constructs an MxGradientCollector and enables training data collection for backpropogation.

Method Summary

Modifier and Type	Method and Description
void	backward(ai.djl.ndarray.NDArray array)
void	close()
static boolean	isRecording() Gets whether Autograd is recording computations.
static boolean	isTraining() Gets whether Autograd is in training/predicting mode.
static boolean	setRecording(boolean isRecording) Sets the status to recording/not recording.
static boolean	setTraining(boolean isTraining) Sets the status to training/predicting.

Methods inherited from class java.lang.Object

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

Constructor Detail

MxGradientCollector

public MxGradientCollector()

Constructs an MxGradientCollector and enables training data collection for backpropogation.

Method Detail

isRecording

public static boolean isRecording()

Gets whether Autograd is recording computations.

Returns:

the current state of recording

isTraining

public static boolean isTraining()

Gets whether Autograd is in training/predicting mode.

Returns:

the current state of training/predicting

setRecording

public static boolean setRecording(boolean isRecording)

Sets the status to recording/not recording. When recording, graph will be constructed for gradient computation.

Parameters:

isRecording - the recording state to be set

Returns:

the previous recording state before this set

setTraining

public static boolean setTraining(boolean isTraining)

Sets the status to training/predicting. This affects ctx.is_train in the device running the operator. For example, Dropout will drop inputs randomly when isTraining=True, while simply passing through if isTraining=False.

Parameters:

isTraining - true if for training

Returns:

the previous status before this set

close

public void close()

Specified by:

close in interface ai.djl.training.GradientCollector

Specified by:

close in interface java.lang.AutoCloseable

backward

public void backward(ai.djl.ndarray.NDArray array)

Specified by:

backward in interface ai.djl.training.GradientCollector