Robust Parameter-Efficient LLM Adaptation on Analog Hardware
This work develops an optimizer-agnostic, parameter-efficient adaptation method based on Low-Rank Adaptation (LoRA), keeping the pretrained weights stored on analog arrays fixed while training the LoRA weights to adapt to downstream tasks and hardware non-idealities.
Jin-Dan Li, Zhao-Xian Wu, Tian-Yi Chen
· 0 citations