ai.djl.nn.recurrent

Class RecurrentBlock

java.lang.Object

ai.djl.nn.AbstractBlock

ai.djl.nn.recurrent.RecurrentBlock

All Implemented Interfaces:

Block

Direct Known Subclasses:

GRU, LSTM, RNN

public abstract class RecurrentBlock
extends AbstractBlock

RecurrentBlock is an abstract implementation of recurrent neural networks.

Recurrent neural networks are neural networks with hidden states. They are very popular for natural language processing tasks, and other tasks which involve sequential data.

This [article](http://karpathy.github.io/2015/05/21/rnn-effectiveness/) written by Andrej Karpathy provides a detailed explanation of recurrent neural networks.

Currently, vanilla RNN, LSTM and GRU are implemented, with both multi-layer and bidirectional support.

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
static class	RecurrentBlock.BaseBuilder<T extends RecurrentBlock.BaseBuilder> The Builder to construct a RecurrentBlock type of Block.

Field Summary

Fields

Modifier and Type	Field and Description
protected boolean	batchFirst
protected boolean	bidirectional
protected float	dropRate
protected int	gates
protected boolean	hasBiases
protected int	numLayers
protected boolean	returnState
protected long	stateSize

Fields inherited from class ai.djl.nn.AbstractBlock

children, inputNames, inputShapes, parameters, parameterShapeCallbacks, version

Constructor Summary

Constructors

Constructor and Description
RecurrentBlock(RecurrentBlock.BaseBuilder<?> builder) Creates a RecurrentBlock object.

Method Summary

Modifier and Type	Method and Description
void	beforeInitialize(Shape[] inputs) Performs any action necessary before initialization.
protected int	getNumDirections()
Shape[]	getOutputShapes(NDManager manager, Shape[] inputs) Returns the expected output shapes of the block for the specified input shapes.
Shape	getParameterShape(java.lang.String name, Shape[] inputShapes) Returns the shape of the specified direct parameter of this block given the shapes of the input to the block.
void	loadMetadata(byte version, java.io.DataInputStream is) Overwrite this to load additional metadata with the parameter values.

Methods inherited from class ai.djl.nn.AbstractBlock

addChildBlock, addParameter, addParameter, addParameter, cast, clear, describeInput, forward, forwardInternal, getChildren, getDirectParameters, getParameters, initialize, initializeChildBlocks, isInitialized, loadParameters, readInputShapes, saveInputShapes, saveMetadata, saveParameters, setInitializer, setInitializer, toString

Methods inherited from class java.lang.Object

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

Methods inherited from interface ai.djl.nn.Block

forward, forward, validateLayout

Field Detail

stateSize

protected long stateSize

dropRate

protected float dropRate

numLayers

protected int numLayers

gates

protected int gates

batchFirst

protected boolean batchFirst

hasBiases

protected boolean hasBiases

bidirectional

protected boolean bidirectional

returnState

protected boolean returnState

Constructor Detail

RecurrentBlock

public RecurrentBlock(RecurrentBlock.BaseBuilder<?> builder)

Creates a RecurrentBlock object.

Parameters:

builder - the Builder that has the necessary configurations

Method Detail

getOutputShapes

public Shape[] getOutputShapes(NDManager manager,
                               Shape[] inputs)

Returns the expected output shapes of the block for the specified input shapes.

Parameters:

manager - an NDManager

inputs - the shapes of the inputs

Returns:

the expected output shapes of the block

beforeInitialize

public void beforeInitialize(Shape[] inputs)

Performs any action necessary before initialization.

Overrides:

beforeInitialize in class AbstractBlock

Parameters:

inputs - the expected shapes of the input

getParameterShape

public Shape getParameterShape(java.lang.String name,
                               Shape[] inputShapes)

Returns the shape of the specified direct parameter of this block given the shapes of the input to the block.

Specified by:

getParameterShape in interface Block

Overrides:

getParameterShape in class AbstractBlock

Parameters:

name - the name of the parameter

inputShapes - the shapes of the input to the block

Returns:

the shape of the parameter specified

loadMetadata

public void loadMetadata(byte version,
                         java.io.DataInputStream is)
                  throws java.io.IOException,
                         MalformedModelException

Overwrite this to load additional metadata with the parameter values.

If you overwrite AbstractBlock.saveMetadata(DataOutputStream) or need to provide backward compatibility to older binary formats, you prabably need to overwrite this. This default implementation checks if the version number fits, if not it throws an MalformedModelException. After that it restores the input shapes.

Overrides:

loadMetadata in class AbstractBlock

Parameters:

version - the version used for loading this metadata.

is - the input stream we are loading from

Throws:

java.io.IOException - loading failed

MalformedModelException - data can be loaded but has wrong format

getNumDirections

protected int getNumDirections()