ai.djl.modality.nlp.embedding

Class TrainableTextEmbedding

java.lang.Object

ai.djl.nn.AbstractBlock

ai.djl.modality.nlp.embedding.TrainableTextEmbedding

All Implemented Interfaces:

TextEmbedding, Block

public class TrainableTextEmbedding
extends AbstractBlock
implements TextEmbedding

TrainableTextEmbedding is an implementation of TextEmbedding based on TrainableWordEmbedding block. This TextEmbedding is ideal when there are no pre-trained embeddings available, or when the pre-trained embedding needs to be further trained.

Field Summary

Fields inherited from class ai.djl.nn.AbstractBlock

children, inputNames, inputShapes, parameters, parameterShapeCallbacks, version

Constructor Summary

Constructors

Constructor and Description
TrainableTextEmbedding(TrainableWordEmbedding wordEmbedding) Constructs a TrainableTextEmbedding.

Method Summary

Modifier and Type	Method and Description
NDArray	embedText(NDArray textIndices) Embeds the text after preprocessed using TextEmbedding.preprocessTextToEmbed(List).
NDList	forward(ParameterStore parameterStore, NDList inputs, boolean training, ai.djl.util.PairList<java.lang.String,java.lang.Object> params) Applies the operating function of the block once.
Shape[]	getOutputShapes(NDManager manager, Shape[] inputShapes) Returns the expected output shapes of the block for the specified input shapes.
void	initializeChildBlocks(NDManager manager, DataType dataType, Shape... inputShapes) Initializes the Child blocks of this block.
long[]	preprocessTextToEmbed(java.util.List<java.lang.String> text) Preprocesses the text to embed into an array to pass into the model.
java.util.List<java.lang.String>	unembedText(NDArray textEmbedding) Returns the closest matching text for a given embedding.

Methods inherited from class ai.djl.nn.AbstractBlock

addChildBlock, addParameter, addParameter, addParameter, beforeInitialize, cast, clear, describeInput, getChildren, getDirectParameters, getParameters, getParameterShape, initialize, isInitialized, loadMetadata, 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.modality.nlp.embedding.TextEmbedding

embedText, embedText

Methods inherited from interface ai.djl.nn.Block

forward, forward, validateLayout

Constructor Detail

TrainableTextEmbedding

public TrainableTextEmbedding(TrainableWordEmbedding wordEmbedding)

Constructs a TrainableTextEmbedding.

Parameters:

wordEmbedding - the word embedding to embed each word

Method Detail

preprocessTextToEmbed

public long[] preprocessTextToEmbed(java.util.List<java.lang.String> text)

Preprocesses the text to embed into an array to pass into the model.

Make sure to call TextEmbedding.embedText(NDManager, long[]) after this.

Specified by:

preprocessTextToEmbed in interface TextEmbedding

Parameters:

text - the text to embed

Returns:

the indices of text that is ready to embed

embedText

public NDArray embedText(NDArray textIndices)
                  throws EmbeddingException

Description copied from interface: TextEmbedding

Embeds the text after preprocessed using TextEmbedding.preprocessTextToEmbed(List).

Specified by:

embedText in interface TextEmbedding

Parameters:

textIndices - the indices of text to embed

Returns:

the embedded text

Throws:

EmbeddingException - if there is an error while trying to embed

unembedText

public java.util.List<java.lang.String> unembedText(NDArray textEmbedding)

Returns the closest matching text for a given embedding.

Specified by:

unembedText in interface TextEmbedding

Parameters:

textEmbedding - the text embedding to find the matching string text for.

Returns:

text similar to the passed in embedding

forward

public NDList forward(ParameterStore parameterStore,
                      NDList inputs,
                      boolean training,
                      ai.djl.util.PairList<java.lang.String,java.lang.Object> params)

Description copied from interface: Block

Applies the operating function of the block once. This method should be called only on blocks that are initialized.

Specified by:

forward in interface Block

Parameters:

parameterStore - the parameter store

inputs - the input NDList

training - true for a training forward pass

params - optional parameters

Returns:

the output of the forward pass

initializeChildBlocks

public void initializeChildBlocks(NDManager manager,
                                  DataType dataType,
                                  Shape... inputShapes)

Description copied from class: AbstractBlock

Initializes the Child blocks of this block. You need to override this method if your subclass has child blocks. Used to determine the correct input shapes for child blocks based on the requested input shape for this block.

Overrides:

initializeChildBlocks in class AbstractBlock

Parameters:

manager - the manager to use for initialization

dataType - the requested data type

inputShapes - the expected input shapes for this block

getOutputShapes

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

Description copied from interface: Block

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

Specified by:

getOutputShapes in interface Block

Parameters:

manager - an NDManager

inputShapes - the shapes of the inputs

Returns:

the expected output shapes of the block