Open-source LLM fine-tuning framework — YAML-config training for QLoRA, FFT, DPO, ORPO on any open-weight model. The community standard for custom models.