Whether you’re doing full fine-tuning, instruction tuning, LoRA/QLoRA, or alignment work, Axolotl makes it easy to get started, without diving deep into training internals. It’s built on top of Hugging Face’s Transformers library but wraps much of the complexity in a clean, YAML-based configuration system. You define your training setup like datasets, models, and training strategy in a single config file. Axolotl takes care of the rest.
Key features:
Axolotl is great for users who want to focus on their data and tasks instead of the details of deep learning internals. With sensible defaults, strong community support, and various integrations, it's a go-to choice for practical fine-tuning of open models.
Last modified 07 May 2026