Rugburn

Software Development

Building Testing Datasets with Hugging Face API

August 30, 2025

This project utilizes the open-source LLM Meta-Llama-3.1-8B-Instruct via the Hugging Face API to create testing datasets in JSON or CSV formats. Key skills include using Transformers, Tokenizers, Quantization for model size reduction, Gradio for UI development, and Google Colab with T4 GPU for enhanced performance.

Llama LLM Engineering

Colab Data Data Science Gradio Hugging Face Multishot Prompting Python Quantization Tokenizer Transformers

Skills

Hugging Face API to load and run open source model
Transformers
- Tokenizer to apply chat template
- Quantization to reduce model size for use on Colab T4 Box.
Gradio for demo UI development
Multishot Prompting to enforce multiple output formats
Google Colab with use of the T4 GPU box

See the source code

More pROJECTS