QLoRA

A PEFT approach that quantizes the base model and trains low‑rank adapters, enabling resource‑efficient fine‑tuning.