Building an Efficiency Pipeline: Commutativity and Cumulativeness of Efficiency Operators for Transformers

TL;DR

This paper investigates the properties of efficiency methods for NLP models, revealing that they are largely commutative and cumulative, which simplifies the design of efficiency pipelines.

Contribution

It demonstrates that multiple efficiency operators can be combined in any order with predictable results, advancing the understanding of efficiency method interactions.

Findings

Efficiency operators are commutative, order has little impact.

Efficiency operators are cumulative, combined effects are predictable.

Guidelines for designing efficiency pipelines are provided.

Abstract

There exists a wide variety of efficiency methods for natural language processing (NLP) tasks, such as pruning, distillation, dynamic inference, quantization, etc. We can consider an efficiency method as an operator applied on a model. Naturally, we may construct a pipeline of multiple efficiency methods, i.e., to apply multiple operators on the model sequentially. In this paper, we study the plausibility of this idea, and more importantly, the commutativity and cumulativeness of efficiency operators. We make two interesting observations: (1) Efficiency operators are commutative -- the order of efficiency methods within the pipeline has little impact on the final results; (2) Efficiency operators are also cumulative -- the final results of combining several efficiency methods can be estimated by combining the results of individual methods. These observations deepen our understanding of efficiency operators and provide useful guidelines for their real-world applications.