AdaFish: Fast low-rank parameter-efficient fine-tuning by using second-order information
This work addresses computational efficiency for researchers and practitioners adapting large models to downstream tasks, though it is incremental as it builds on existing low-rank fine-tuning frameworks.
The paper tackles the problem of slow training in parameter-efficient fine-tuning of large pretrained models by introducing AdaFish, a second-order algorithm that leverages low-rank structure in the Fisher information matrix, achieving competitive performance with AdamW in numerical experiments.
Recent advancements in large-scale pretrained models have significantly improved performance across a variety of tasks in natural language processing and computer vision. However, the extensive number of parameters in these models necessitates substantial memory and computational resources for full training. To adapt these models for downstream tasks or specific application-oriented datasets, parameter-efficient fine-tuning methods leveraging pretrained parameters have gained considerable attention. However, it can still be time-consuming due to lots of parameters and epochs. In this work, we introduce AdaFish, an efficient algorithm of the second-order type designed to expedite the training process within low-rank decomposition-based fine-tuning frameworks. Our key observation is that the associated generalized Fisher information matrix is either low-rank or extremely small-scaled. Such a generalized Fisher information matrix is shown to be equivalent to the Hessian matrix. Moreover, we prove the global convergence of AdaFish, along with its iteration/oracle complexity. Numerical experiments show that our algorithm is quite competitive with the state-of-the-art AdamW method.