Active Budget Allocation for Efficient Scaling Law Estimation via Surrogate-Guided Pruning

TL;DR

This paper introduces a surrogate-guided pruning method combined with Successive Halving to efficiently estimate scaling laws, significantly reducing computational costs while improving accuracy.

Contribution

It proposes a novel approach that integrates surrogate models with Successive Halving for more efficient and accurate scaling law estimation.

Findings

Surrogate-guided Successive Halving outperforms naive methods in learning curve prediction.

Achieves up to 98.7% reduction in computational costs compared to exhaustive approaches.

Improves the accuracy of scaling law estimation with mean relative gains up to 5.47%.

Abstract

Predicting model performance at larger scales enables the design of training strategies and architectures tailored to specific performance targets. Empirical scaling law research identifies functional forms to aid this prediction task. These describe the relationship between loss and compute using a loss-compute frontier defined by learning curves. Due to the empirical nature of this approach, the computational burden is substantial, making strategic resource allocation essential - yet it remains surprisingly underexplored. In this work, we address this shortcoming by exploring the suitability of Successive Halving (SH) and SH combined with parametric and non-parametric surrogate models. In addition to enabling a more systematic allocation of a given compute budget, our findings show that SH paired with surrogate models yields a set of learning curves that includes one with a lower loss-compute value than what naive uniform allocation or an SH-only approach can obtain. Our experiments demonstrate mean relative improvements of up to 2.84% and 5.47% on real-world and synthetic learning curve datasets. This strategic resource allocation enables us to obtain accurate scaling laws at significantly reduced computational costs, saving up to 98.7% over the traditional exhaustive approach.