LoRA (Low-Rank Adaptation) [LoRA]

LoRA reduces computational costs by adapting only small parts of a model during fine-tuning.