SlickTune 🧩 Documentation¶
🧠 SlickTune 🧩 Philosophy¶
SlickTune is SlickML’s composable toolkit for fine-tuning large language models. Fine-tuning is treated as an orthogonal stack —
model × strategy × objective × data × metrics
— so you can swap LoRA for DoRA / AdaLoRA / QLoRA / full FT without rewriting the rest. Built on Transformers + PEFT + TRL, with probes and holdout metrics so you can verify the model actually learned your facts.
Requirements and how to set up your Python environment with uv …
Train LoRA SFT, probe facts, and run holdout eval in a few commands …
Visual guide to Full FT, LoRA, DoRA, AdaLoRA, and QLoRA for beginners …
Explore the SlickTune API and source modules …
Stay up-to-date with new features and fixes …
🧑💻🤝 Become a Contributor¶
SlickTune is building an open-source community for practical LLM fine-tuning. Development details live in our Contributing guidelines. Special thanks to all contributors 👇
❓ 🆘 📲 Need Help?¶
Please join our Slack Channel to interact with the core team and community, or email admin@slickml.com.