Grams: Gradient Descent with Adaptive Momentum Scaling
This addresses the challenge of efficient training and fine-tuning for large language models, representing a novel method rather than an incremental improvement.
The paper tackles the problem of optimizing deep learning models by introducing Grams, an algorithm that decouples direction and magnitude in parameter updates, achieving faster convergence and better generalization compared to state-of-the-art optimizers like Adam and Lion.
We introduce $\mathbf{G}$radient Descent with $\mathbf{A}$daptive $\mathbf{M}$omentum $\mathbf{S}$caling ($\mathbf{Grams}$), a novel optimization algorithm that decouples the direction and magnitude of parameter updates in deep learning. Unlike traditional optimizers that directly integrate momentum into updates, Grams separates the update direction, derived from current gradients, from momentum, which is used solely for adaptive magnitude scaling. This approach enables Grams to achieve improved loss descent compared to state-of-the-art cautious and momentum-based optimizers. We theoretically demonstrate that Grams descents faster than other state-of-the-art optimizers and establish a global convergence guarantee for Grams. We also validate its effectiveness through extensive empirical evaluations. The results demonstrate Grams' superior performance, including faster convergence and better generalization, compared to widely-used optimizers such as Adam, Lion, and their cautious variants. Our results highlight Grams' potential as a transformative approach for efficiently training and fine-tuning large language models. Code is available at https://github.com/Gunale0926/Grams.