Memory-Efficient Adaptive Optimization

TL;DR

This paper introduces a memory-efficient adaptive optimization method that maintains the benefits of per-parameter adaptivity, enabling training of larger models and batch sizes with faster convergence and reduced memory usage.

Contribution

The paper proposes a novel adaptive optimizer with significantly lower memory overhead while preserving convergence guarantees and effectiveness in large-scale language model training.

Findings

Achieves up to 2-fold speedups in training large models

Reduces memory usage compared to traditional optimizers

Enables training larger models and batch sizes

Abstract

Adaptive gradient-based optimizers such as Adagrad and Adam are crucial for achieving state-of-the-art performance in machine translation and language modeling. However, these methods maintain second-order statistics for each parameter, thus introducing significant memory overheads that restrict the size of the model being used as well as the number of examples in a mini-batch. We describe an effective and flexible adaptive optimization method with greatly reduced memory overhead. Our method retains the benefits of per-parameter adaptivity while allowing significantly larger models and batch sizes. We give convergence guarantees for our method, and demonstrate its effectiveness in training very large translation and language models with up to 2-fold speedups compared to the state-of-the-art.