1 paper
Jui-Nan Yen, Si Si, Zhao Meng +5
Low-rank adaption (LoRA) is a widely used parameter-efficient finetuning method for LLM that reduces memory requirements. However, current LoRA optimizers lack transformation invar…