Scaling Laws for Upcycling Mixture-of-Experts Language Models

Seng Pei Liew, Takuya Kato, Sho Takase

Keywords: language modeling, mixture of experts, scaling law, upcycling

TL;DR: Exploring scaling laws for upcycling dense language models to MoE, revealing key trade-offs and guidelines for efficient training.

Abstract:

Pretraining large language models (LLMs) is resource-intensive, often requiring months of training time even with high-end GPU clusters. There are two approaches of mitigating such computational demands: reusing smaller models to train larger ones (upcycling), and training computationally efficient models like mixture-of-experts (MoE). In this paper, we study the upcycling of LLMs to MoE models, of which the scaling behavior remains underexplored. Through extensive experiments, we identify empirical scaling laws that describe how performance depends on dataset size and model configuration. Particularly, we show that, while scaling these factors improves performance, there is a novel interaction term between the dense and upcycled training dataset that limits the efficiency of upcycling at large computational budgets.
Based on these findings, we provide guidance to scale upcycling, and establish conditions under which upcycling outperforms from-scratch trainings within budget constraints.

Submission Number: 22