ai.djl.nn.core

Class Embedding.BaseBuilder<T,B extends Embedding.BaseBuilder<T,B>>

java.lang.Object

ai.djl.nn.core.Embedding.BaseBuilder<T,B>

Type Parameters:

T - the type of object to embed

Direct Known Subclasses:

TrainableWordEmbedding.Builder

Enclosing class:

Embedding<T>

public abstract static class Embedding.BaseBuilder<T,B extends Embedding.BaseBuilder<T,B>>
extends java.lang.Object

The Builder to construct a Embedding type of Block.

Field Summary

Fields

Modifier and Type	Field and Description
protected DataType	dataType
protected T	defaultItem
protected int	embeddingSize
protected java.lang.Class<T>	embeddingType
protected AbstractIndexedEmbedding<T>	fallthrough
protected boolean	sparseGrad
protected boolean	useDefault

Constructor Summary

Constructors

Modifier	Constructor and Description
protected	BaseBuilder()

Method Summary

Modifier and Type	Method and Description
java.lang.Class<T>	getEmbeddingType() Returns the embedded type.
B	optDataType(DataType dataType) Sets the data type of the embedding arrays (default is Float32).
B	optDefaultItem(T defaultItem) Sets whether to use a default item's embedding for undefined items.
B	optFallthrough(AbstractIndexedEmbedding<T> fallthrough) Sets a custom handler for items not found in the embedding.
B	optSparseGrad(boolean sparseGrad) Sets the optional parameter whether to compute row sparse gradient in the backward calculation.
B	optUseDefault(boolean useDefault) Sets whether to use a default embedding for undefined items (default true).
protected abstract B	self() Returns this {code Builder} object.
B	setEmbeddingSize(int embeddingSize) Sets the size of the embeddings.
protected abstract B	setType(java.lang.Class<T> embeddingType) Creates a new Embedding.BaseBuilder with the specified embedding type.

Methods inherited from class java.lang.Object

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

Field Detail

embeddingType

protected java.lang.Class<T> embeddingType

embeddingSize

protected int embeddingSize

useDefault

protected boolean useDefault

defaultItem

protected T defaultItem

fallthrough

protected AbstractIndexedEmbedding<T> fallthrough

sparseGrad

protected boolean sparseGrad

dataType

protected DataType dataType

Constructor Detail

BaseBuilder

protected BaseBuilder()

Method Detail

getEmbeddingType

public java.lang.Class<T> getEmbeddingType()

Returns the embedded type.

Returns:

the embedded type

setType

protected abstract B setType(java.lang.Class<T> embeddingType)

Creates a new Embedding.BaseBuilder with the specified embedding type.

Parameters:

embeddingType - the embedding class

Returns:

a new Embedding.BaseBuilder class with the specified embedding type

setEmbeddingSize

public B setEmbeddingSize(int embeddingSize)

Sets the size of the embeddings.

Parameters:

embeddingSize - the size of the 1D embedding array

Returns:

this Builder

optUseDefault

public B optUseDefault(boolean useDefault)

Sets whether to use a default embedding for undefined items (default true).

Parameters:

useDefault - true to provide a default embedding and false to throw an IllegalArgumentException when the item can not be found

Returns:

this Builder

optDefaultItem

public B optDefaultItem(T defaultItem)

Sets whether to use a default item's embedding for undefined items.

Parameters:

defaultItem - the item to use as a default.

Returns:

this Builder

optFallthrough

public B optFallthrough(AbstractIndexedEmbedding<T> fallthrough)

Sets a custom handler for items not found in the embedding.

See the standard fallthrough handlers optUseDefault(boolean) and optDefaultItem(Object).

Parameters:

fallthrough - the embedding to handle default cases.

Returns:

this Builder

optSparseGrad

public B optSparseGrad(boolean sparseGrad)

Sets the optional parameter whether to compute row sparse gradient in the backward calculation. If set to True, the grad’s storage type is row_sparse.

Parameters:

sparseGrad - whether to compute row sparse gradient in the backward calculation

Returns:

this Builder

optDataType

public B optDataType(DataType dataType)

Sets the data type of the embedding arrays (default is Float32).

Parameters:

dataType - the dataType to use for the embedding

Returns:

this Builder

self

protected abstract B self()

Returns this {code Builder} object.

Returns:

this BaseBuilder