Efficient Large Scale Language Modeling with Mixtures of Experts
This work addresses the challenge of computational efficiency in large-scale language modeling for AI researchers and practitioners, though it is incremental as it builds on existing MoE concepts with empirical validation.
The paper tackles the problem of scaling language models efficiently by comparing Mixture of Experts (MoE) layers with dense models across various tasks, finding that MoEs are substantially more compute-efficient, matching dense model performance with about 4 times less compute at modest budgets and outperforming compute-equivalent dense models at scale.
Mixture of Experts layers (MoEs) enable efficient scaling of language models through conditional computation. This paper presents a detailed empirical study of how autoregressive MoE language models scale in comparison with dense models in a wide range of settings: in- and out-of-domain language modeling, zero- and few-shot priming, and full-shot fine-tuning. With the exception of fine-tuning, we find MoEs to be substantially more compute efficient. At more modest training budgets, MoEs can match the performance of dense models using $\sim$4 times less compute. This gap narrows at scale, but our largest MoE model (1.1T parameters) consistently outperforms a compute-equivalent dense model (6.7B parameters). Overall, this performance gap varies greatly across tasks and domains, suggesting that MoE and dense models generalize differently in ways that are worthy of future study. We make our code and models publicly available for research use.