Model Merging in Pre-training of Large Language Models

TL;DR

This paper investigates model merging techniques during large-scale pre-training of language models, demonstrating performance gains, cost reductions, and providing practical guidelines based on extensive experiments with various architectures.

Contribution

It offers a comprehensive analysis of model merging during pre-training, revealing its benefits and underlying mechanisms, and introduces practical guidelines for implementation.

Findings

Merging checkpoints trained with constant learning rates improves performance.

Model merging enables accurate prediction of annealing behavior.

Significant reduction in training costs and improved efficiency.

Abstract

Model merging has emerged as a promising technique for enhancing large language models, though its application in large-scale pre-training remains relatively unexplored. In this paper, we present a comprehensive investigation of model merging techniques during the pre-training process. Through extensive experiments with both dense and Mixture-of-Experts (MoE) architectures ranging from millions to over 100 billion parameters, we demonstrate that merging checkpoints trained with constant learning rates not only achieves significant performance improvements but also enables accurate prediction of annealing behavior. These improvements lead to both more efficient model development and significantly lower training costs. Our detailed ablation studies on merging strategies and hyperparameters provide new insights into the underlying mechanisms while uncovering novel applications. Through comprehensive experimental analysis, we offer the open-source community practical pre-training guidelines for effective model merging.