LoRA-based Adaptation

Fine-tuning a model using Low-Rank Adaptation, a parameter-efficient method that adds small trainable layers to a frozen base model.