Class HuggingFaceInference

Class implementing the Large Language Model (LLM) interface using the Hugging Face Inference API for text generation.

Hierarchy

  • LLM
    • HuggingFaceInference

Implements

Constructors

Properties

CallOptions: BaseLLMCallOptions
ParsedCallOptions: Omit<BaseLLMCallOptions, never>
apiKey: undefined | string = undefined

API key to use.

caller: AsyncCaller

The async caller should be used by subclasses to make any async calls, which will thus benefit from the concurrency and retry logic.

endpointUrl: undefined | string = undefined

Custom inference endpoint URL to use

frequencyPenalty: undefined | number = undefined

Penalizes repeated tokens according to frequency

maxTokens: undefined | number = undefined

Maximum number of tokens to generate in the completion.

model: string = "gpt2"

Model to use

temperature: undefined | number = undefined

Sampling temperature to use

topK: undefined | number = undefined

Integer to define the top tokens considered within the sample operation to create new text.

topP: undefined | number = undefined

Total probability mass of tokens to consider at each step

verbose: boolean

Whether to print out response text.

callbacks?: Callbacks
metadata?: Record<string, unknown>
tags?: string[]

Accessors

Methods

  • This method takes an input and options, and returns a string. It converts the input to a prompt value and generates a result based on the prompt.

    Parameters

    Returns Promise<string>

    A string result based on the prompt.

  • This method is similar to call, but it's used for making predictions based on the input text.

    Parameters

    • text: string

      Input text for the prediction.

    • Optional options: string[] | BaseLLMCallOptions

      Options for the LLM call.

    • Optional callbacks: Callbacks

      Callbacks for the LLM call.

    Returns Promise<string>

    A prediction based on the input text.

  • Stream all output from a runnable, as reported to the callback system. This includes all inner runs of LLMs, Retrievers, Tools, etc. Output is streamed as Log objects, which include a list of jsonpatch ops that describe how the state of the run has changed in each step, and the final state of the run. The jsonpatch ops can be applied in order to construct state.

    Parameters

    Returns AsyncGenerator<RunLogPatch, any, unknown>

  • Default implementation of transform, which buffers input and then calls stream. Subclasses should override this method if they can start producing output while input is still being generated.

    Parameters

    Returns AsyncGenerator<string, any, unknown>

Generated using TypeDoc