This project provides a fine-tuned Llama2 model that enables code generation, execution, debugging, and internet access for users needing to automate complex tasks or analyze data programmatically. It aims to empower developers and researchers by integrating LLM capabilities with a functional code execution environment.

How It Works

The model is fine-tuned from CodeLlama-7B-Instruct, enhancing its ability to generate, identify, and execute code blocks. It maintains state by monitoring and retaining Python variables across executed code snippets, facilitating iterative development and complex data manipulation. This approach allows the LLM to act as an interactive coding assistant, directly executing generated code and using the results for subsequent operations.

Quick Start & Requirements

• Install dependencies: pip install -r requirements.txt
• Run the Gradio App: python3 chatbot.py --path Seungyoun/codellama-7b-instruct-pad
• Requires Python 3.x.
• Further details and alternative model paths are available in the README.

Highlighted Details

• Achieved 70.12 pass@1 on HumanEval benchmarks, significantly outperforming the base CodeLlama-7B-Instruct.
• Demonstrates capabilities in fetching stock data via Yahoo Finance and plotting with Matplotlib.
• Supports variable persistence across code execution sessions.
• Focus on data development for GPT-4 code interpretation and model enhancement.

Maintenance & Community

• Contributions, issues, and feature requests are welcome via the issues page.
• The project acknowledges contributions from llama2 and yet-another-gpt-tutorial repositories.

Licensing & Compatibility

• Distributed under the MIT License.
• Permissive license suitable for commercial use and integration into closed-source projects.

Limitations & Caveats

The project is actively focused on data development for GPT-4 code interpretation, suggesting ongoing development and potential for breaking changes. Specific details on supported languages beyond Python and frameworks are not explicitly detailed in the README.