Package me.damoebe.architectures.transformer.embedding

Class Sequence

java.lang.Object

me.damoebe.architectures.transformer.embedding.Sequence

public class Sequence
extends Object

A container class for a transformer internal Embedding-Sequence

Constructor Summary

Constructors

Constructor	Description
Sequence()	Creates a new empty Embedding-Sequence.
Sequence(List<Double> rawSequenceData, int tokenSize)	Constructs a Sequence from a raw-input-data list by tokenizing it and positional-encode each token.
Sequence(List<Embedding> embeddings)	Constructs a Sequence from an embedding list (used for sub-sequence splitting)

Method Summary

Modifier and Type	Method	Description
List<Embedding>	embeddings()	Getter for the stored embeddings
double[][]	getHorizontalMatrix()	Gets the horizontal matrix of this sequence (rows -> embeddingData, columns -> embedding)
List<Double>	getRawList()	Generate a 'raw' Double list for this sequence containing all embedding values
List<Sequence>	getSubSequences(int subSequenceAmount)	Splits this sequences into n sub-sequences
double[][]	getVerticalMatrix()	Gets the vertical (standard dimensions) matrix of this sequence where rows -> embeddings and columns -> embeddingData

Methods inherited from class java.lang.Object

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

Constructor Details

Sequence

public Sequence()

Creates a new empty Embedding-Sequence.

Sequence

public Sequence(List<Double> rawSequenceData,
 int tokenSize)

Constructs a Sequence from a raw-input-data list by tokenizing it and positional-encode each token.

Parameters:

rawSequenceData - The input sequence (un-tokenized) as a Double List.

tokenSize - The data-size each token (embedding) should have.

Sequence

public Sequence(List<Embedding> embeddings)

Constructs a Sequence from an embedding list (used for sub-sequence splitting)

Parameters:

embeddings - The List of Embeddings from which this Sequence should be constructed

Method Details

getRawList

public List<Double> getRawList()

Generate a 'raw' Double list for this sequence containing all embedding values

Returns:

The raw Double list

getHorizontalMatrix

public double[][] getHorizontalMatrix()

Gets the horizontal matrix of this sequence (rows -> embeddingData, columns -> embedding)

Returns:

The horizontal Matrix

getVerticalMatrix

public double[][] getVerticalMatrix()

Gets the vertical (standard dimensions) matrix of this sequence where rows -> embeddings and columns -> embeddingData

Returns:

The vertical Matrix

getSubSequences

public List<Sequence> getSubSequences(int subSequenceAmount)

Splits this sequences into n sub-sequences

Parameters:

subSequenceAmount - The amount of sub-sequences this sequence should be split into

Returns:

A list of all sub-sequences of this sequence

embeddings

public List<Embedding> embeddings()

Getter for the stored embeddings

Returns:

A list with all embeddings