Kernelized Sparse Fine-Tuning with Bi-level Parameter Competition for Vision Models

TL;DR

This paper introduces SNELLA, a novel one-stage sparse fine-tuning method for vision models that improves task relevance detection, increases adaptation capacity through kernelized low-rank updates, and reduces memory usage while achieving state-of-the-art results.

Contribution

SNELLA combines kernelized low-rank updates with adaptive bi-level sparsity for efficient, high-performance vision model fine-tuning in a single stage.

Findings

SNELLA outperforms previous methods on classification, segmentation, and generation tasks.

Achieves 1.8% higher Top-1 accuracy on FGVC benchmark.

Reduces memory usage by 31.1%-39.9% across various models.

Abstract

Parameter-efficient fine-tuning (PEFT) aims to adapt pre-trained vision models to downstream tasks. Among PEFT paradigms, sparse tuning achieves remarkable performance by adjusting only the weights most relevant to downstream tasks, rather than densely tuning the entire weight matrix. Current methods follow a two-stage paradigm. First, it locates task-relevant weights by gradient information, which overlooks the parameter adjustments during fine-tuning and limits the performance. Second, it updates only the located weights by applying a sparse mask to the gradient of the weight matrix, which results in high memory usage due to the storage of all weight matrices in the optimizer. In this paper, we propose a one-stage method named SNELLA to overcome the above limitations. For memory usage, SNELLA selectively updates the weight matrix by adding it to another sparse matrix that is merged by two low-rank learnable matrices. We extend the low-rank decomposition by introducing nonlinear kernel functions, thereby increasing the rank of the resulting merged matrix to prevent the interdependency among weight updates, enabling better adaptation to downstream tasks. For locating task-relevant weights, we propose an adaptive bi-level sparsity allocation mechanism that encourages weights to compete across and inside layers based on their importance scores in an end-to-end manner. Extensive experiments are conducted on classification, segmentation, and generation tasks using different pre-trained vision models. The results show that SNELLA achieves SOTA performance with low memory usage. Notably, SNELLA obtains 1.8% (91.9% v.s. 90.1%) higher Top-1 accuracy on the FGVC benchmark compared to SPT-LoRA. Compared to previous methods, SNELLA achieves a memory reduction of 31.1%-39.9% across models with parameter scales from 86M to 632M. Our source codes are available at https://github.com/ssfgunner/SNELL.