Parameter-Efficient Fine-Tuning via Circular Convolution
This addresses the problem of balancing efficiency and performance in parameter-efficient fine-tuning for AI practitioners, though it appears incremental as it builds on LoRA.
The paper tackles the performance limitations of Low-Rank Adaptation (LoRA) in fine-tuning large models by proposing Circular Convolution Adaptation (C^3A), which achieves high-rank adaptation and outperforms LoRA and its variants across various tasks.
Low-Rank Adaptation (LoRA) has gained popularity for fine-tuning large foundation models, leveraging low-rank matrices $\mathbf{A}$ and $\mathbf{B}$ to represent weight changes (i.e., $Δ\mathbf{W} = \mathbf{B} \mathbf{A}$). This method reduces trainable parameters and mitigates heavy memory consumption associated with full delta matrices by sequentially multiplying $\mathbf{A}$ and $\mathbf{B}$ with the activation. Despite its success, the intrinsic low-rank characteristic may limit its performance. Although several variants have been proposed to address this issue, they often overlook the crucial computational and memory efficiency brought by LoRA. In this paper, we propose Circular Convolution Adaptation (C$^3$A), which not only achieves high-rank adaptation with enhanced performance but also excels in both computational power and memory utilization. Extensive experiments demonstrate that C$^3$A consistently outperforms LoRA and its variants across various fine-tuning tasks.