Joint MoE Scaling Laws: Mixture of Experts Can Be Memory Efficient
This work addresses the problem of efficient large-scale machine learning model deployment for researchers and practitioners, offering insights into designing and deploying MoE models in practical scenarios.
This work tackled the problem of Mixture of Experts (MoE) model scalability under memory constraints, resulting in the finding that MoE models can be more memory-efficient than dense models, with experiments involving up to 2.7B active parameters. The researchers derived joint scaling laws for dense and MoE models, providing a framework for optimal MoE configuration selection.
Mixture of Experts (MoE) architectures have significantly increased computational efficiency in both research and real-world applications of large-scale machine learning models. However, their scalability and efficiency under memory constraints remain relatively underexplored. In this work, we present joint scaling laws for dense and MoE models, incorporating key factors such as the number of active parameters, dataset size, and the number of experts. Our findings provide a principled framework for selecting the optimal MoE configuration under fixed memory and compute budgets. Surprisingly, we show that MoE models can be more memory-efficient than dense models, contradicting conventional wisdom. To derive and validate the theoretical predictions of our scaling laws, we conduct over 280 experiments with up to 2.7B active parameters and up to 5B total parameters. These results offer actionable insights for designing and deploying MoE models in practical large-scale training scenarios.