This repository demonstrates the fine-tuning of the Llama-3-8B model and compares its performance with the base model. It is part of a comparative study between fine-tuning and Retrieval-Augmented Generation (RAG) to determine which approach is more suitable for our use case.

The detailed blog can be found here.

Fine_tuning.ipynb contains all the code necessary for fine-tuning the model.

For this project, we will be using publicly available medical data. This dataset is structured as prompt-completion pairs, where users ask medical questions and receive relevant responses from doctors. (Data Source)

You can access the fine-tuned model on Hugging Face via this link.

For questions or feedback about the project, don't hesitate to reach out to me on LinkedIn.

The RAG implementation for this study can be found here.