transformer/transformer library

Classes

EmbeddingLayer
A layer that turns a 1D vector of integer indices into a 2D matrix of dense vectors.
EmbeddingLayerMatrix
A layer that turns a 2D matrix of integer indices into a 3D tensor of dense vectors.
GlobalAveragePooling1D
A Global Average Pooling layer for 1D data.
LayerNormalization
A Layer Normalization layer for 2D Matrix data.
LayerNormalizationVector
A Layer Normalization layer for 1D Vector data.
MultiHeadAttention
Implements the Multi-Head Self-Attention mechanism.
PositionalEncoding
Injects information about the relative or absolute position of tokens in a sequence.
TransformerEncoderBlock
Implements a single Transformer Encoder Block.

Functions

layerNorm(Tensor<Matrix> m, Tensor<Vector> gamma, Tensor<Vector> beta, {double epsilon = 1e-5}) Tensor<Matrix>
layerNormVector(Tensor<Vector> v, Tensor<Vector> gamma, Tensor<Vector> beta, {double epsilon = 1e-5}) Tensor<Vector>