Class EmbeddingLookup

Namespace

NeuralNetworks.Operations.Parameterized

Assembly

NeuralNetworks.dll

Embedding lookup operation that maps integer indices to embedding vectors.

public class EmbeddingLookup : ParamOperation<int[,], float[,], float[,]>, IParamOperation

Inheritance

object

Operation

Operation<int[,], float[,]>

ParamOperation<int[,], float[,]>

ParamOperation<int[,], float[,], float[,]>

EmbeddingLookup

Implements

IParamOperation

Inherited Members

ParamOperation<int[,], float[,], float[,]>.Param

ParamOperation<int[,], float[,], float[,]>.Backward(float[,])

ParamOperation<int[,], float[,], float[,]>.CalcParamGradient(float[,])

ParamOperation<int[,], float[,], float[,]>.GetData()

ParamOperation<int[,], float[,], float[,]>.ApplyData(ParamOperationData, int, int)

ParamOperation<int[,], float[,], float[,]>.GetParamCount()

ParamOperation<int[,], float[,], float[,]>.UpdateParams(Optimizer)

ParamOperation<int[,], float[,]>.GetParamCount()

ParamOperation<int[,], float[,]>.UpdateParams(Optimizer)

ParamOperation<int[,], float[,]>.GetData()

ParamOperation<int[,], float[,]>.ApplyData(ParamOperationData, int, int)

Operation<int[,], float[,]>.Input

Operation<int[,], float[,]>.Output

Operation<int[,], float[,]>.Forward(int[,], bool)

Operation<int[,], float[,]>.Backward(float[,])

Operation<int[,], float[,]>.CalcOutput(bool)

Operation<int[,], float[,]>.CalcInputGradient(float[,])

Operation<int[,], float[,]>.Forward(object, bool)

Operation<int[,], float[,]>.Backward(object)

Operation<int[,], float[,]>.GetOutputType()

Operation<int[,], float[,]>.GetInputType()

Operation.SetRegistered()

Operation.GetOutputType()

Operation.GetInputType()

Operation.Forward(object, bool)

Operation.Backward(object)

object.Equals(object)

object.Equals(object, object)

object.GetHashCode()

object.GetType()

object.MemberwiseClone()

object.ReferenceEquals(object, object)

Remarks

Initializes a new instance of the EmbeddingLookup class.

Constructors

EmbeddingLookup(float[,])

Embedding lookup operation that maps integer indices to embedding vectors.

public EmbeddingLookup(float[,] embeddings)

Parameters

embeddings float[,]

Embedding matrix [vocabSize, embeddingDim].

Remarks

Initializes a new instance of the EmbeddingLookup class.

Methods

CalcInputGradient(float[,])

Calculates the gradient of the input indices based on the output gradient.

protected override int[,] CalcInputGradient(float[,] outputGradient)

Parameters

outputGradient float[,]

Returns

int[,]

Input gradient (not used for integer indices, returns zero array).

CalcOutput(bool)

protected override float[,] CalcOutput(bool inference)

Parameters

inference bool

Returns

float[,]

Embeddings [batchSize, sequenceLength, embeddingDim] flattened to [batchSize, sequenceLength * embeddingDim].

CalcParamGradient(float[,])

Calculates the gradient of the embedding parameters based on the output gradient.

protected override float[,] CalcParamGradient(float[,] outputGradient)

Parameters

outputGradient float[,]

Gradient from upstream [batchSize, sequenceLength * embeddingDim].

Returns

float[,]

ToString()

Returns a string that represents the current operation.

public override string ToString()

Returns

string

A string that represents the current operation.

Remarks

It is used in a layer/model description.