private-gpt/llama2-code-interpreter

Make Llama2 use Code Execution, Debug, Save Code, Reuse it, Access to Internet

Llama2 Code Interpreter

🤗 CodeLlama 7B Finetuned Model (HF)

This project allows LLM to generate code, execute it, receive feedback, debug, and answer questions based on the whole process. It is designed to be intuitive and versatile, capable of dealing with multiple languages and frameworks.

The purpose and direction of the project

Quick Start

Run the Gradio App:

python3 chatbot.py --path Seungyoun/codellama-7b-instruct-pad

News

• 🔥🔥🔥[2023/08/27] We're thrilled to announce that our 🤗 Llama2 Code Interpreter-7B (Finetuned from CodeLlama-7B-Instruct) model achieved a remarkable 70.12pass@1 on the HumanEval Benchmarks.

HumanEval

Model	Score(pass@1)
Codellama instruct 7b	34.8%
Codellama instruct 7b - finetuning	70.12%

GSM8K

Model	Score
Code Llama 7B	13%
Code Llama 13B	20.8%
Codellama instruct 7b - finetuning	28%

🌟 Key Features

• 🚀 Code Generation and Execution: Llama2 is capable of generating code, which it then automatically identifies and executes within its generated code blocks.
• Monitors and retains Python variables that were used in previously executed code blocks.
• 🌟 At the moment, my focus is on "Data development for GPT-4 code interpretation" and "Enhancing the model using this data". For more details, check out the feat/finetuning branch in our repository.
• 🌟 CodeLlama Support CodeLlama2

Examples

---

Llama2 in Action

In the GIF, Llama2 is seen in action. A user types in the request: Plot Nvidia 90 days chart. Llama2, an advanced code interpreter fine-tuned on a select dataset, swiftly queries Yahoo Finance. Moments later, it fetches the latest Nvidia stock prices from the past 90 days. Using Matplotlib, Llama2 then generates a clear and detailed stock price chart for Nvidia, showcasing its performance over the given period.

Installation

1. Clone the Repository (if you haven't already):

   git clone https://github.com/SeungyounShin/Llama2-Code-Interpreter.git
   cd Llama2-Code-Interpreter

2. Install the required dependencies:

   pip install -r requirements.txt

---

Run App with GPT4 finetunned Llama Model

To start interacting with Llama2 via the Gradio UI using codellama-7b-instruct-pad, follow the steps below:

2. Run the Gradio App:

   python3 chatbot.py --path Seungyoun/codellama-7b-instruct-pad

For those who want to use other models:

General Instructions to Run App

To start interacting with Llama2 via the Gradio UI using other models:

1. Run the Command:

   python3 chatbot.py --model_path <your-model-path>

Replace <your-model-path> with the path to the model file you wish to use. A recommended model for chat interactions is meta-llama/Llama-2-13b-chat.

Contributions

Contributions, issues, and feature requests are welcome! Feel free to check issues page.

License

Distributed under the MIT License. See LICENSE for more information.

Contact

Seungyoun, Shin - [email protected]

Acknowledgement

Here are some relevant and related projects that have contributed to the development of this work:

1. llama2 : GitHub Repository
2. yet-another-gpt-tutorial : GitHub Repository

These projects have been instrumental in providing valuable insights and resources, and their contributions are highly appreciated.

---