RAG

This project designed for question-answering tasks using the Retrieval-Augmented Generation (RAG) approach. It leverages a pre-trained language model (ollama) and a retrieval model (ColBERTv2) to generate accurate answers to a variety of questions. The module includes functionality to easily configure and use alternative datasets, such as SQuAD, for diverse question-answering needs.

Features

• RAG Model Integration: Combines retrieval and generation models for enhanced answer accuracy.
• Configurable Retrieval: Adjust the number of passages retrieved to optimize context relevance.
• Chain of Thought Generation: Uses a chain of thought process to generate coherent answers.
• Dataset Flexibility: Easily switch between different datasets (e.g., SQuAD, HotPotQA) for diverse question-answering needs.

Usage

1. Installation: Install dependencies using pip install -r requirements.txt.
2. Configure Server: Ensure the ollama server is running locally.
3. Load Dataset: Modify rag_model.py to use the desired dataset (default is SQuAD).
4. Run Module: Execute python rag_model.py to check server, configure DSPy, and ask sample questions.