Class TextEmbedding

Class to handle text embedding model on top of onnxruntime

Hierarchy (view full)

Constructors

constructor

Properties

eos feed sess?

Methods

argmax embed initializeFeed load release updateKVCache

Constructors

constructor

Properties

Protectedeos

eos: bigint = 2n

Protectedfeed

feed: Record<string, Tensor> = {}

Protected Optionalsess

sess?: InferenceSession

Methods

Protectedargmax

embed

  • embed(tokens): Promise<Float32Array<ArrayBufferLike>>

  • Generate embeddings from input tokens

    Parameters

    • tokens: bigint[]

      Input tokens to generate embeddings from

    Returns Promise<Float32Array<ArrayBufferLike>>

    Float32Array containing the embedding vector

initializeFeed

load

  • load(model, onnx_file?, options): Promise<void>

  • Parameters

    • model: string
    • onnx_file: string = "onnx/model.onnx"
    • options: LoadOptions

    Returns Promise<void>

release

ProtectedupdateKVCache

Settings

Member Visibility

On This Page

Constructors
Properties
Methods