https://x.com/BerivanISIK/status/1825577160486302044
Super excited about new work Lottery Ticket Adaptation (LoTA): https://t.co/GMViXLtDqo
— Berivan Isik (@BerivanISIK) August 19, 2024
We propose a sparse adaptation method that finetunes only a sparse subset of the pre-trained weights. LoTA mitigates catastrophic forgetting and enables model merging by breaking the… https://t.co/cPHBxrUaQc pic.twitter.com/XGxquoISPp
현직자들 있으면 설명좀
이건 continuous learning이 아님... catastrophic forgetting하곤 아무상관도 없음... model merge 쪽이라고 봐야함
그닥