Could Model Partitioning Make Federated Learning More Sustainable?

2026-08-14

AI Summary

This study addresses the escalating carbon footprint and excessive energy consumption of high-carbon clients in federated learning by proposing a novel carbon-aware model splitting method. Innovatively integrating model splitting into sustainable federated learning, this approach enables adaptive training task scheduling through dynamic computation offloading to alternative nodes in response to real-time carbon signals. Experimental results demonstrate that, under specific splitting strategies, energy consumption at high-carbon clients is reduced by up to 76% without introducing significant overhead. By effectively balancing system performance with environmental sustainability, this work establishes an efficient computation offloading paradigm for green federated learning, offering a practical solution to mitigate the ecological impact of distributed machine learning systems while maintaining training efficacy.

Abstract

As federated learning (FL) extends from distributed machine learning between low-power devices to cross-silo scenarios involving edge servers and data centres, its carbon footprint has become a growing concern. Addressing this, methods for sustainable FL align training with low-carbon energy availability or low grid demand and reduce the energy consumption of clients powered by high-carbon sources by decreasing the size of their models. We propose applying model partitioning, which can shift energy consumption by offloading parts of a model to another participant, in response to carbon- or grid-aware signals. Our preliminary findings show that for some partition points, model partitioning can reduce a participant's energy consumption by up to 76% without any significant time or energy consumption overhead compared to non-partitioned training.

Problem

• Federated Learning
• Sustainability
• Carbon Footprint
• Energy Consumption

Innovation

• Model Partitioning
• Sustainable Federated Learning
• Carbon-aware Computing
• Energy Offloading