Coeff-Tuning: A Graph Filter Subspace View for Tuning Attention-Based Large Models
This work addresses the challenge of efficiently customizing large models for downstream tasks, offering a novel approach that improves performance while maintaining low computational costs, though it is incremental as it builds on existing parameter-efficient fine-tuning methods.
The paper tackles the problem of parameter-efficient fine-tuning for large pre-trained transformers by proposing Coeff-Tuning, which learns a small set of coefficients to construct an expressive filter subspace from multi-head attention maps, achieving superior performance with negligible additional parameters compared to existing methods.
Transformer-based large pre-trained models have shown remarkable generalization ability, and various parameter-efficient fine-tuning (PEFT) methods have been proposed to customize these models on downstream tasks with minimal computational and memory budgets. Previous PEFT methods are primarily designed from a tensor-decomposition perspective that tries to effectively tune the linear transformation by finding the smallest subset of parameters to train. Our study adopts an orthogonal view by representing the attention operation as a graph convolution and formulating the multi-head attention maps as a convolutional filter subspace, with each attention map as a subspace element. In this paper, we propose to tune the large pre-trained transformers by learning a small set of combination coefficients that construct a more expressive filter subspace from the original multi-head attention maps. We show analytically and experimentally that the tuned filter subspace can effectively expand the feature space of the multi-head attention and further enhance the capacity of transformers. We further stabilize the fine-tuning with a residual parameterization of the tunable subspace coefficients, and enhance the generalization with a regularization design by directly applying dropout on the tunable coefficient during training. The tunable coefficients take a tiny number of parameters and can be combined with previous PEFT methods in a plug-and-play manner. Extensive experiments show that our approach achieves superior performances than PEFT baselines with neglectable additional parameters.