ai.djl.nn.recurrent

Class RecurrentBlock.BaseBuilder<T extends RecurrentBlock.BaseBuilder>

java.lang.Object

ai.djl.nn.recurrent.RecurrentBlock.BaseBuilder<T>

Direct Known Subclasses:

GRU.Builder, LSTM.Builder, RNN.Builder

Enclosing class:

RecurrentBlock

public abstract static class RecurrentBlock.BaseBuilder<T extends RecurrentBlock.BaseBuilder>
extends java.lang.Object

The Builder to construct a RecurrentBlock type of Block.

Field Summary

Fields

Modifier and Type	Field and Description
protected RNN.Activation	activation
protected boolean	batchFirst
protected boolean	bidirectional
protected float	dropRate
protected boolean	hasBiases
protected int	numLayers
protected boolean	returnState
protected long	stateSize

Constructor Summary

Constructors

Constructor and Description
BaseBuilder()

Method Summary

Modifier and Type	Method and Description
T	optBatchFirst(boolean batchFirst) Sets the optional batchFirst flag that indicates whether the input is batch major or not.
T	optBidirectional(boolean useBidirectional) Sets the optional parameter that indicates whether to use bidirectional recurrent layers.
T	optDropRate(float dropRate) Sets the drop rate of the dropout on the outputs of each RNN layer, except the last layer.
T	optHasBiases(boolean hasBiases) Sets the optional biases flag that indicates whether to use biases or not.
T	optReturnState(boolean returnState) Sets the optional flag that indicates whether to return state or not.
protected abstract T	self()
T	setNumLayers(int numLayers) Sets the Required number of stacked layers.
T	setStateSize(int stateSize) Sets the Required size of the state for each layer.

Methods inherited from class java.lang.Object

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

Field Detail

dropRate

protected float dropRate

stateSize

protected long stateSize

numLayers

protected int numLayers

batchFirst

protected boolean batchFirst

hasBiases

protected boolean hasBiases

bidirectional

protected boolean bidirectional

returnState

protected boolean returnState

activation

protected RNN.Activation activation

Constructor Detail

BaseBuilder

public BaseBuilder()

Method Detail

optDropRate

public T optDropRate(float dropRate)

Sets the drop rate of the dropout on the outputs of each RNN layer, except the last layer.

Parameters:

dropRate - the drop rate of the dropout

Returns:

this Builder

setStateSize

public T setStateSize(int stateSize)

Sets the Required size of the state for each layer.

Parameters:

stateSize - the size of the state for each layer

Returns:

this Builder

setNumLayers

public T setNumLayers(int numLayers)

Sets the Required number of stacked layers.

Parameters:

numLayers - the number of stacked layers

Returns:

this Builder

optBidirectional

public T optBidirectional(boolean useBidirectional)

Sets the optional parameter that indicates whether to use bidirectional recurrent layers.

Parameters:

useBidirectional - whether to use bidirectional recurrent layers

Returns:

this Builder

optBatchFirst

public T optBatchFirst(boolean batchFirst)

Sets the optional batchFirst flag that indicates whether the input is batch major or not. The default value is true.

Parameters:

batchFirst - whether the input is batch major or not

Returns:

this Builder

optHasBiases

public T optHasBiases(boolean hasBiases)

Sets the optional biases flag that indicates whether to use biases or not.

Parameters:

hasBiases - whether to use biases or not

Returns:

this Builder

optReturnState

public T optReturnState(boolean returnState)

Sets the optional flag that indicates whether to return state or not. This is typically useful when you use RecurrentBlock in Sequential block. The default value is false.

Parameters:

returnState - whether to return state or not

Returns:

this Builder

self

protected abstract T self()