Exploring Learngene via Stage-wise Weight Sharing for Initializing Variable-sized Models

TL;DR

This paper introduces SWS, a stage-wise weight sharing method for initializing variable-sized models efficiently, leveraging learngene layers learned from large models to reduce training costs and improve performance.

Contribution

The paper proposes a novel SWS approach that enhances learngene-based initialization by incorporating stage guidance, significantly reducing training costs and storage for variable-sized models.

Findings

SWS outperforms models trained from scratch on ImageNet-1K.

Reduces training costs by approximately 6.6x.

Achieves better results with minimal fine-tuning after 1 epoch.

Abstract

In practice, we usually need to build variable-sized models adapting for diverse resource constraints in different application scenarios, where weight initialization is an important step prior to training. The Learngene framework, introduced recently, firstly learns one compact part termed as learngene from a large well-trained model, after which learngene is expanded to initialize variable-sized models. In this paper, we start from analysing the importance of guidance for the expansion of well-trained learngene layers, inspiring the design of a simple but highly effective Learngene approach termed SWS (Stage-wise Weight Sharing), where both learngene layers and their learning process critically contribute to providing knowledge and guidance for initializing models at varying scales. Specifically, to learn learngene layers, we build an auxiliary model comprising multiple stages where the layer weights in each stage are shared, after which we train it through distillation. Subsequently, we expand these learngene layers containing stage information at their corresponding stage to initialize models of variable depths. Extensive experiments on ImageNet-1K demonstrate that SWS achieves consistent better performance compared to many models trained from scratch, while reducing around 6.6x total training costs. In some cases, SWS performs better only after 1 epoch tuning. When initializing variable-sized models adapting for different resource constraints, SWS achieves better results while reducing around 20x parameters stored to initialize these models and around 10x pre-training costs, in contrast to the pre-training and fine-tuning approach.