This repository offers tools for fine-tuning causal language models, leveraging the power of the PEFT (PyTorch Efficient Fine-Tuning) library. It's designed for enhancing code generation tailored to specific repositories.
- Model Flexibility 📈: Load any causal language model.
- Configuration-Based Setup ⚙️: Easily define model selection and tokenization through a
conf/config.yamlfile. - Efficient Fine-Tuning 🏃♂️: Currently utilizing LoRA (Low-Rank Adaptation) via the PEFT library. More methods coming soon!
- Future-Ready 🔮: Stay tuned for training methods for instruct and QA models.
- Python 3.x 🐍
- PyTorch 🔥
- PEFT Library 📚
- Other dependencies listed in
requirements.txt
- Clone the repository:
git clone [repository URL]:git: - Make a place for storing data
mkdir data📂 make sure to updateconfig.yamlaccordingly. - Install required packages:
pip install -r requirements.txt
Tweak the conf\config.yaml file to set your model and tokenization parameters.
Kick off the fine-tuning process by running the main script (python main.py). Fine tuned model will be saved in directory specified in config.yaml -> "optimization" -> result_dir.
We welcome your contributions to make this tool even more efficient and feature-rich. Please adhere to standard open-source contribution guidelines.
Apache License 2.0