GRIN: GRadient-INformed MoE

TL;DR

GRIN introduces a gradient-informed training method for MoE models, enabling efficient sparse routing and improved scaling, resulting in models that outperform or match larger dense models across various tasks.

Contribution

The paper presents GRIN, a novel training approach for MoE models that incorporates sparse gradient estimation and optimized parallelism, enhancing scalability and performance.

Findings

A 6.6B parameter MoE model surpasses a 7B dense model on multiple benchmarks.

GRIN enables training of large MoE models without token dropping.

Extensive evaluations show significant performance improvements across diverse tasks.

Abstract

Mixture-of-Experts (MoE) models scale more effectively than dense models due to sparse computation through expert routing, selectively activating only a small subset of expert modules. However, sparse computation challenges traditional training practices, as discrete expert routing hinders standard backpropagation and thus gradient-based optimization, which are the cornerstone of deep learning. To better pursue the scaling power of MoE, we introduce GRIN (GRadient-INformed MoE training), which incorporates sparse gradient estimation for expert routing and configures model parallelism to avoid token dropping. Applying GRIN to autoregressive language modeling, we develop a top-2 16$\times$3.8B MoE model. Our model, with only 6.6B activated parameters, outperforms a 7B dense model and matches the performance of a 14B dense model trained on the same data. Extensive evaluations across diverse tasks demonstrate the potential of GRIN to significantly enhance MoE efficacy, achieving 79.4 on MMLU, 83.7 on HellaSwag, 74.4 on HumanEval, and 58.9 on MATH.