REVISITING LARS FOR LARGE BATCH TRAINING GENERALIZATION OF NEURAL NETWORKS

TL;DR — Analysis on LARS, LAMB and new optimizers for general applications and large-batch training

Abstract

LARS and LAMB have emerged as prominent techniques in Large Batch Learn- ing (LBL), ensuring the stability of AI training. One of the primary challenges in LBL is convergence stability, where the AI agent usually gets trapped into the sharp minimizer. Addressing this challenge, a relatively recent technique, known as warm-up, has been employed. However, warm-up lacks a strong theoretical foundation, leaving the door open for further exploration of more efficacious al- gorithms. In light of this situation, we conduct empirical experiments to analyze the behaviors of the two most popular optimizers in the LARS family: LARS and LAMB, with and without a warm-up strategy. Our analyses give a compre- hensive insight into the behavior of LARS, LAMB, and the necessity of a warm- up technique in LBL, including an explanation of their failure in many cases. Building upon these insights, we propose a novel algorithm called Time Varying LARS (TVLARS), which facilitates robust training in the initial phase without the need for warm-up. We run extensive experimental evaluations to demonstrate that TVLARS achieves competitive results with LARS and LAMB when warm-up is utilized while surpassing their performance without the warm-up technique.