posembeddings

Documentation

Index

• type Config
• type LearnedPositionalEmbeddings
  • func NewLearnedPositionalEmbeddings(config Config) *LearnedPositionalEmbeddings
  • func (m *LearnedPositionalEmbeddings) Encode(positions []int) []ag.Node

Types

type Config

type Config struct {
	NumEmbeddings int
	EmbeddingDim  int
	PaddingIDX    int
	Offset        int
}

Config provides configuration settings for a LearnedPositionalEmbeddings Model.

type LearnedPositionalEmbeddings

type LearnedPositionalEmbeddings struct {
	nn.BaseModel
	Config  Config
	Vectors []nn.Param `spago:"type:weights;scope:model"`
}

LearnedPositionalEmbeddings contains positional embeddings fine-tuned during the training phase.

func NewLearnedPositionalEmbeddings

func NewLearnedPositionalEmbeddings(config Config) *LearnedPositionalEmbeddings

NewLearnedPositionalEmbeddings returns a new LearnedPositionalEmbeddings. TODO: PaddingIDX

func (*LearnedPositionalEmbeddings) Encode

func (m *LearnedPositionalEmbeddings) Encode(positions []int) []ag.Node

Encode performs the forward step for each input and returns the result.