Timezone: »

Sparse Structure Search for Delta Tuning
Shengding Hu · Zhen Zhang · Ning Ding · Yadao Wang · Yasheng Wang · Zhiyuan Liu · Maosong Sun

Adapting large pre-trained models (PTMs) through fine-tuning imposes prohibitive computational and storage burdens. Recent studies of delta tuning (DT), i.e., parameter-efficient tuning, find that only optimizing a small portion of parameters conditioned on PTMs could yield on-par performance compared to conventional fine-tuning. Generally, DT methods exquisitely design delta modules (DT modules) which could be applied to arbitrary fine-grained positions inside PTMs. However, the effectiveness of these fine-grained positions largely relies on sophisticated manual designation, thereby usually producing sub-optimal results. In contrast to the manual designation, we explore constructing DT modules in an automatic manner. We automatically \textbf{S}earch for the \textbf{S}parse \textbf{S}tructure of \textbf{Delta} Tuning (S$^3$Delta). Based on a unified framework of various DT methods, S$^3$Delta conducts the differentiable DT structure search through bi-level optimization and proposes shifted global sigmoid method to explicitly control the number of trainable parameters. Extensive experiments show that S$^3$Delta surpasses manual and random structures with less trainable parameters. The searched structures preserve more than 99\% fine-tuning performance with 0.01\% trainable parameters. Moreover, the advantage of S$^3$Delta is amplified with extremely low trainable parameters budgets (0.0009\%$\sim$0.01\%). The searched structures are transferable and explainable, providing suggestions and guidance for the future design of DT methods. Our codes are publicly available at \url{https://github.com/thunlp/S3Delta}.

Author Information

Shengding Hu (Tsinghua University)
Zhen Zhang (Tsinghua University, Tsinghua University)
Ning Ding (Tsinghua University, Tsinghua University)
Yadao Wang (Huawei Technologies Ltd.)
Yasheng Wang (Huawei Noah's Ark Lab)
Zhiyuan Liu (Tsinghua University)
Maosong Sun (Tsinghua University)

More from the Same Authors