Group and Shuffle: Efficient Structured Orthogonal Parametrization
This work addresses the need for more efficient fine-tuning methods in large-scale AI models, offering incremental improvements over existing orthogonal fine-tuning approaches.
The paper tackles the problem of efficient fine-tuning of large neural networks by introducing a new class of structured matrices that generalizes previous work, which improves parameter and computational efficiency in orthogonal fine-tuning. It validates the method empirically on text-to-image diffusion models and language modeling tasks, showing gains in efficiency.
The increasing size of neural networks has led to a growing demand for methods of efficient fine-tuning. Recently, an orthogonal fine-tuning paradigm was introduced that uses orthogonal matrices for adapting the weights of a pretrained model. In this paper, we introduce a new class of structured matrices, which unifies and generalizes structured classes from previous works. We examine properties of this class and build a structured orthogonal parametrization upon it. We then use this parametrization to modify the orthogonal fine-tuning framework, improving parameter and computational efficiency. We empirically validate our method on different domains, including adapting of text-to-image diffusion models and downstream task fine-tuning in language modeling. Additionally, we adapt our construction for orthogonal convolutions and conduct experiments with 1-Lipschitz neural networks.