Decoupled Multi-Predictor Optimization for Inference-Efficient Model Tuning

TL;DR

This paper introduces DMPO, a novel method for optimizing multi-predictor models to improve inference efficiency by decoupling feature representation and discrimination in early stages, leading to better performance with less computation.

Contribution

The paper proposes a decoupled optimization framework and lightweight architecture enhancements for multi-stage predictors to improve inference efficiency in large-scale pre-trained models.

Findings

DMPO outperforms existing methods in reducing computational cost.

Decoupling improves the quality of early-stage features.

Two-phase loss weighting enhances model training effectiveness.

Abstract

Recently, remarkable progress has been made in large-scale pre-trained model tuning, and inference efficiency is becoming more crucial for practical deployment. Early exiting in conjunction with multi-stage predictors, when cooperated with a parameter-efficient fine-tuning strategy, offers a straightforward way to achieve an inference-efficient model. However, a key challenge remains unresolved: How can early stages provide low-level fundamental features to deep stages while simultaneously supplying high-level discriminative features to early-stage predictors? To address this problem, we propose a Decoupled Multi-Predictor Optimization (DMPO) method to effectively decouple the low-level representative ability and high-level discriminative ability in early stages. First, in terms of architecture, we introduce a lightweight bypass module into multi-stage predictors for functional decomposition of shallow features from early stages, while a high-order statistics-based predictor is developed for early stages to effectively enhance their discriminative ability. To reasonably train our multi-predictor architecture, a decoupled optimization is proposed to allocate two-phase loss weights for multi-stage predictors during model tuning, where the initial training phase enables the model to prioritize the acquisition of discriminative ability of deep stages via emphasizing representative ability of early stages, and the latter training phase drives discriminative ability towards earlier stages as much as possible. As such, our DMPO can effectively decouple representative and discriminative abilities in early stages in terms of architecture design and model optimization. Experiments across various datasets and pre-trained backbones demonstrate that DMPO clearly outperforms its counterparts when reducing computational cost.