LAMP: Look-Ahead Mixed-Precision Inference of Large Language Models

TL;DR

This paper introduces LAMP, an adaptive mixed-precision inference method for large language models that selectively increases accuracy in critical components, significantly improving inference precision with minimal recomputation.

Contribution

The paper presents a novel adaptive strategy for mixed-precision computation in transformer inference, optimizing accuracy while reducing computational overhead.

Findings

Up to two orders of magnitude accuracy improvement on GPT-2 models.

Very low recomputation rates achieve significant precision gains.

Adaptive component selection enhances transformer inference efficiency.

Abstract

Mixed-precision computations are a hallmark of the current stage of AI, driving the progress in large language models towards efficient, locally deployable solutions. This article addresses the floating-point computation of compositionally-rich functions, concentrating on transformer inference. Based on the rounding error analysis of a composition $f(g(\mathrm{x}))$, we provide an adaptive strategy that selects a small subset of components of $g(\mathrm{x})$ to be computed more accurately while all other computations can be carried out with lower accuracy. We then explain how this strategy can be applied to different compositions within a transformer and illustrate its overall effect on transformer inference. We study the effectiveness of this algorithm numerically on GPT-2 models and demonstrate that already very low recomputation rates allow for improvements of up to two orders of magnitude in accuracy.