Code Generation with LLMs

Description

This repository contains the code for our final year project at Yildiz Technical University.

You can find the models in the following links in huggingfaces' model hub:

• codegen-instruct on HF hub
• mistral-code on HF hub
• phi2-code on HF hub

Table of Contents

• Installation
• Usage
• Contributing
• License
• Contact

Installation

This repository comes with a utility package parsakzr/codellm

To install and use the project locally, follow these steps:

1. Clone the repository

git clone [repo-url]
cd [repo-name]

2. Install the dependencies

pip install -r requirements.txt

Optional: Also, if you want to run the notebooks, you need to install jupyter notebook.

pip install jupyter

Usage

The directory structure of the project is as follows:

.
├── finetuning
│   ├── codegen
│   ├── humaneval
│   └── mbpp
├── evaluation
│   ├── evalanalysis.ipynb
│   ├── CodeGen
│   ├── mistral-code
│   ├── phi2-code
│   └── etc.
└── GUI
    ├── EvalModelOnGradio.ipynb
    ├── run.py
    └── etc.

• In the finetuning directory, you can find the code for the finetuning of the models. just run the notebooks in Colab, or locally.
• In the evaluation directory, you can find the code for the evaluation of the models.
• In the GUI directory, you can find the code to run the GUI. You can run the GUI by running the run.py file. or use the notebook to run it in Colab.

To run the UI

cd GUI
pip install -r requirements.txt
python run.py

And then go to given URL in your browser.

Contributing

Contributions are always welcome! Please create a Pull Request to contribute. If you find any bugs, please report them as issues.

License

This project is licensed under the MIT License.

Contact

• GitHub: parsakzr

References

• Code Generation with LLMs
• EvalPlus
• CodeAlpaca Dataset