ChatHuggingFace

​

Node Details

• Name: ChatHuggingFace
• Type: ChatHuggingFace
• Version: 3.0
• Category: Chat Models

​

Base Classes

• ChatHuggingFace
• BaseChatModel
• HuggingFaceInference (and its base classes)

​

Parameters

​

Credential (Required)

​

Parameters

​

Credential (Required)

• Type: huggingFaceApi

​

Inputs

1. Model (Required)

​

Inputs

1. Model (Required)

   • Type: string
   • Description: The name of the HuggingFace model to use (e.g., “gpt2”)
   • Note: Leave blank if using a custom inference endpoint

2. Cache (Optional)

3. Cache (Optional)

   • Type: BaseCache
   • Description: Caching mechanism for the model

4. Endpoint (Optional)

5. Endpoint (Optional)

   • Type: string
   • Description: Custom inference endpoint URL
   • Example: https://xyz.eu-west-1.aws.endpoints.huggingface.cloud/gpt2

6. Temperature (Optional)

7. Temperature (Optional)

   • Type: number
   • Description: Controls randomness in output generation
   • Step: 0.1

8. Max Tokens (Optional)

9. Max Tokens (Optional)

   • Type: number
   • Description: Maximum number of tokens to generate
   • Step: 1

10. Top Probability (Top P) (Optional)

11. Top Probability (Top P) (Optional)

    • Type: number
    • Description: Nucleus sampling parameter
    • Step: 0.1

12. Top K (Optional)

13. Top K (Optional)

    • Type: number
    • Description: Top-K sampling parameter
    • Step: 0.1

14. Frequency Penalty (Optional)

15. Frequency Penalty (Optional)

    • Type: number
    • Description: Penalizes frequent token usage
    • Step: 0.1

16. Stop Sequence (Optional)

17. Stop Sequence (Optional)

    • Type: string
    • Description: Comma-separated list of sequences to stop generation

​

Initialization

The node initializes a HuggingFaceInference instance with the provided parameters. It handles parameter parsing and credential retrieval.

​

Usage

This node is particularly useful for:

1. Integrating HuggingFace models into chat applications
2. Customizing language model behavior for specific use cases
3. Experimenting with different model parameters
4. Using custom-trained or fine-tuned models via custom endpoints

​

Notes

• Some parameters may not apply to all models. Users should check the specific model’s documentation for compatible parameters.
• The node supports both HuggingFace’s hosted models and custom inference endpoints, providing flexibility in deployment options.