This study is great and addresses a question I've had about LoRA for a while. In a continual learning paper from last year, I found LoRA was extremely effectiv

This study is great and addresses a question I've had about LoRA for a while. In a continual learning paper from last year, I found LoRA was extremely effective for faster fine-tuning and not forgetting the original dataset: https://arxiv.org/abs/2306.01904 https://arxiv.org/abs/2306.01904