SARA: Singular-Value Based Adaptive Low-Rank Adaption
This work addresses the need for adaptive rank selection in fine-tuning for users of large models, offering an incremental improvement over existing LoRA methods.
The paper tackles the problem of manually tuning rank values in LoRA for parameter-efficient fine-tuning of large pre-trained models, proposing SARA and Mo-SARA methods that adaptively determine optimal ranks per layer using SVD, resulting in effective and parameter-efficient performance across various complex tasks.
With the increasing number of parameters in large pre-trained models, LoRA as a parameter-efficient fine-tuning(PEFT) method is widely used for not adding inference overhead. The LoRA method assumes that weight changes during fine-tuning can be approximated by low-rank matrices. However, the rank values need to be manually verified to match different downstream tasks, and they cannot accommodate the varying importance of different layers in the model. In this work, we first analyze the relationship between the performance of different layers and their ranks using SVD. Based on this, we design the Singular-Value Based Adaptive Low-Rank Adaption(SARA), which adaptively finds the rank during initialization by performing SVD on the pre-trained weights. Additionally, we explore the Mixture-of-SARA(Mo-SARA), which significantly reduces the number of parameters by fine-tuning only multiple parallel sets of singular values controlled by a router. Extensive experiments on various complex tasks demonstrate the simplicity and parameter efficiency of our methods. They can effectively and adaptively find the most suitable rank for each layer of each model.