NeuralFinGPT

NeuralFinGPT is a comprehensive project aimed at downloading, preprocessing, and analyzing financial reports for MENA companies using NLP techniques. This repository includes code for creating a PostgreSQL database, fine-tuning the Mistral7b LLM using QLoRA, and implementing RAG algorithms for context creation to answer specific financial queries.

Table of Contents

• Introduction
• Features
• Installation
• Usage
• Project Structure
• Contributing
• License

Introduction

NeuralFinGPT is designed to facilitate advanced financial data analysis for the MENA region by leveraging state-of-the-art NLP techniques and large language models.

Features

• Download and preprocess financial reports.
• Create and manage a PostgreSQL database.
• Fine-tune Mistral7b LLM using QLoRA.
• Implement RAG algorithms for context-based queries.

Installation

1. Clone the repository:

   git clone https://github.com/ranystephan/NeuralFinGPT.git

2. Navigate to the project directory:

   cd NeuralFinGPT

3. Install the required dependencies:

   pip install -r requirements.txt

Usage

1. Run the data collection script:

   jupyter notebook data_collection.ipynb

2. Fine-tune the model using:

   jupyter notebook fine-tuning_RAG_func.ipynb

3. Analyze financial data and run queries.

Project Structure

• data_collection.ipynb: Script for downloading and preprocessing financial reports.
• fine-tuning_RAG_func.ipynb: Notebook for fine-tuning the language model.
• neuralfingpt-draft1.pdf: Initial project draft and documentation.

Contributing

Contributions are welcome! Please submit a pull request or open an issue to discuss your ideas.

License

This project is licensed under the MIT License.