HF TGI LLM Connector

The Text Generation Inference is a Rust, Python, and gRPC server specifically designed for text generation inference. It can be self-hosted to power LLM APIs and inference widgets.

This node can connect to locally or remotely hosted TGI servers which includes Text Generation Inference Endpoints of popular text generation models that are deployed via Hugging Face Hub.

For more details and information about integrating with the Hugging Face TextGen Inference and setting up a local server, refer to the LangChain documentation.

Options

Hugging Face TextGen Inference Server Settings

Inference Server URL

The URL of the inference server to use, e.g. http://localhost:8010/.

Model Parameters

Seed

Set the seed parameter to any integer of your choice and use the same value across requests to have reproducible outputs. The default value of 0 means that no seed is specified.

Top k

The number of top-k tokens to consider when generating text.

Typical p

The typical probability threshold for generating text.

Repetition penalty

The repetition penalty to use when generating text.

Max new tokens

The maximum number of tokens to generate in the completion.

The token count of your prompt plus max new tokens cannot exceed the model's context length.

Number of concurrent requests

Maximum number of concurrent requests to LLMs that can be made, whether through API calls or to an inference server. Exceeding this limit may result in temporary restrictions on your access.

It is important to plan your usage according to the model provider's rate limits, and keep in mind that both software and hardware constraints can impact performance.

For OpenAI, please refer to the Limits page for the rate limits available to you.

Temperature

Sampling temperature to use, between 0.0 and 100.0. Higher values will make the output more random, while lower values will make it more focused and deterministic.

Top-p sampling

An alternative to sampling with temperature, where the model considers the results of the tokens (words) with top_p probability mass. Hence, 0.1 means only the tokens comprising the top 10% probability mass are considered.

Input Ports

Icon

An optional Hugging Face hub connection that can be used to access protected Hugging Face inference endpoints.

Output Ports

Icon

Connection to an LLM hosted on a TGI server.

Popular Predecessors

  • No recommendations found

Popular Successors

  • No recommendations found

Views

This node has no views

Workflows

  • No workflows found

Links

Developers

You want to see the source code for this node? Click the following button and we’ll use our super-powers to find it for you.