Towards the Systematic Reporting of the Energy and Carbon Footprints of\n Machine Learning

Abstract

Accurate reporting of energy and carbon usage is essential for understanding\nthe potential climate impacts of machine learning research. We introduce a\nframework that makes this easier by providing a simple interface for tracking\nrealtime energy consumption and carbon emissions, as well as generating\nstandardized online appendices. Utilizing this framework, we create a\nleaderboard for energy efficient reinforcement learning algorithms to\nincentivize responsible research in this area as an example for other areas of\nmachine learning. Finally, based on case studies using our framework, we\npropose strategies for mitigation of carbon emissions and reduction of energy\nconsumption. By making accounting easier, we hope to further the sustainable\ndevelopment of machine learning experiments and spur more research into energy\nefficient algorithms.\n