Pear: Pruning and Sharing Adapters in Visual Parameter-Efficient Fine-Tuning

TL;DR

Pear is a novel adapter-pruning framework that enhances the efficiency of visual model fine-tuning by pruning, sharing, and preserving adapter information, leading to improved performance and reduced redundancy.

Contribution

The paper introduces a new adapter-pruning method with sharing and knowledge checkpoint strategies for more efficient visual model adaptation.

Findings

Outperforms existing methods on visual adaptation benchmarks.

Reduces storage and computational costs significantly.

Maintains or improves fine-tuning performance.

Abstract

Adapters have been widely explored to alleviate computational and storage costs when fine-tuning pretrained foundation models. However, the adapter itself can exhibit redundancy, leading to unnecessary storage overhead and inferior performance. In this paper, we propose Prune and Share (Pear), a novel adapter-pruning framework for efficient fine-tuning of pretrained visual foundation models. Specifically, we prune certain adapters and share the more important unpruned ones with positions where adapters are pruned, allowing continual adaptation at these positions after pruning. Additionally, a knowledge checkpoint strategy is introduced, which preserves the information of the pruned adapters and further boosts performance. Experimental results on visual adaptation benchmark validate the effectiveness and efficiency of the proposed Pear comparing to other competitive methods. Code is in https://github.com/yibozhong/pear.