Practical tradeoffs between memory, compute, and performance in learned optimizers
This addresses the high computational and memory costs of learned optimizers for machine learning practitioners, offering an incremental improvement.
The paper tackles the trade-offs between memory, compute, and performance in learned optimizers, quantifying design features and constructing a new optimizer that is faster and more memory efficient than prior work.
Optimization plays a costly and crucial role in developing machine learning systems. In learned optimizers, the few hyperparameters of commonly used hand-designed optimizers, e.g. Adam or SGD, are replaced with flexible parametric functions. The parameters of these functions are then optimized so that the resulting learned optimizer minimizes a target loss on a chosen class of models. Learned optimizers can both reduce the number of required training steps and improve the final test loss. However, they can be expensive to train, and once trained can be expensive to use due to computational and memory overhead for the optimizer itself. In this work, we identify and quantify the design features governing the memory, compute, and performance trade-offs for many learned and hand-designed optimizers. We further leverage our analysis to construct a learned optimizer that is both faster and more memory efficient than previous work. Our model and training code are open source.