Nhảy tới nội dung

Class: HuggingFaceEmbedding

Uses feature extraction from '@xenova/transformers' to generate embeddings. Per default the model XENOVA_ALL_MINILM_L6_V2 is used.

Can be changed by setting the modelType parameter in the constructor, e.g.:

new HuggingFaceEmbedding({
    modelType: HuggingFaceEmbeddingModelType.XENOVA_ALL_MPNET_BASE_V2,
});

Extends​

Constructors​

new HuggingFaceEmbedding()​

new HuggingFaceEmbedding(init?): HuggingFaceEmbedding

Parameters​

• init?: Partial <HuggingFaceEmbedding>

Returns​

HuggingFaceEmbedding

Overrides​

BaseEmbedding . constructor

Source​

packages/core/src/embeddings/HuggingFaceEmbedding.ts:28

Properties​

embedBatchSize​

embedBatchSize: number = DEFAULT_EMBED_BATCH_SIZE

Inherited from​

BaseEmbedding . embedBatchSize

Source​

packages/core/src/embeddings/types.ts:21

embedInfo?​

optional embedInfo: EmbeddingInfo

Inherited from​

BaseEmbedding . embedInfo

Source​

packages/core/src/embeddings/types.ts:22

extractor​

private extractor: any

Source​

packages/core/src/embeddings/HuggingFaceEmbedding.ts:26

modelType​

modelType: string = HuggingFaceEmbeddingModelType.XENOVA_ALL_MINILM_L6_V2

Source​

packages/core/src/embeddings/HuggingFaceEmbedding.ts:23

quantized​

quantized: boolean = true

Source​

packages/core/src/embeddings/HuggingFaceEmbedding.ts:24

Methods​

getExtractor()​

getExtractor(): Promise<any>

Returns​

Promise<any>

Source​

packages/core/src/embeddings/HuggingFaceEmbedding.ts:33

getQueryEmbedding()​

getQueryEmbedding(query): Promise<null | number[]>

Parameters​

• query: MessageContentDetail

Returns​

Promise<null | number[]>

Inherited from​

BaseEmbedding . getQueryEmbedding

Source​

packages/core/src/embeddings/types.ts:34

getTextEmbedding()​

getTextEmbedding(text): Promise<number[]>

Parameters​

• text: string

Returns​

Promise<number[]>

Overrides​

BaseEmbedding . getTextEmbedding

Source​

packages/core/src/embeddings/HuggingFaceEmbedding.ts:43

getTextEmbeddings()​

getTextEmbeddings(texts): Promise<number[][]>

Optionally override this method to retrieve multiple embeddings in a single request

Parameters​

• texts: string[]

Returns​

Promise<number[][]>

Inherited from​

BaseEmbedding . getTextEmbeddings

Source​

packages/core/src/embeddings/types.ts:48

getTextEmbeddingsBatch()​

getTextEmbeddingsBatch(texts, options?): Promise<number[][]>

Get embeddings for a batch of texts

Parameters​

• texts: string[]

• options?

• options.logProgress?: boolean

Returns​

Promise<number[][]>

Inherited from​

BaseEmbedding . getTextEmbeddingsBatch

Source​

packages/core/src/embeddings/types.ts:64

similarity()​

similarity(embedding1, embedding2, mode): number

Parameters​

• embedding1: number[]

• embedding2: number[]

• mode: SimilarityType= SimilarityType.DEFAULT

Returns​

number

Inherited from​

BaseEmbedding . similarity

Source​

packages/core/src/embeddings/types.ts:24

transform()​

transform(nodes, _options?): Promise <BaseNode <Metadata>[]>

Parameters​

• nodes: BaseNode <Metadata>[]

• _options?: any

Returns​

Promise <BaseNode <Metadata>[]>

Inherited from​

BaseEmbedding . transform

Source​

packages/core/src/embeddings/types.ts:78

truncateMaxTokens()​

truncateMaxTokens(input): string[]

Parameters​

• input: string[]

Returns​

string[]

Inherited from​

BaseEmbedding . truncateMaxTokens

Source​

packages/core/src/embeddings/types.ts:90