Carbontracker is a tool for tracking and predicting the carbon footprint and energy expenditure of Deep Learning (DL) training.
DL has become a powerful tool for achieving impressive results across various tasks. Still, it requires extensive training on specialized hardware accelerators, resulting in an energy-intensive workload that could contribute significantly to climate change.
The growing carbon footprint of the AI industry motivates the search for and development of tools that can measure and reduce the environmental impact of AI.
Carbontracker uses Machine Learning (ML) models to estimate the carbon emissions associated with different hardware and software configurations and predict future training runs' carbon footprint. It also supports predicting the total duration, energy, and carbon footprint of training a DL model. Using the supported APIs, the carbon intensity of electricity production during the expected time is forecasted. Then, the carbon intensity is used to indicate the carbon footprint.
Carbontracker is one of the tools developed as part of the SustainML project to address the environmental impact of Machine Learning and promote sustainability in this field. This project, SustainML, is funded by the European Union to develop a sustainable ML framework. The project aims to address the environmental impact of ML and make ML more energy-efficient and sustainable by developing new algorithms, models, and tools.
SustainML also promotes ML's ethical and social responsibility, including fairness, transparency, and privacy. The project involves collaboration between multiple universities, research institutions, and private companies across Europe. It aims to create a community of researchers, developers, and users committed to sustainable ML.
By Raghavendra Selvan from Machine Learning Section, Kiehn Lab & Data Science Lab, University of Copenhagen.
All the SustainML participants: DFKI, Inria, IBM, University of Copenhagen, UpMem, Technische Universität Kaiserslautern and eProsima.
MORE INFORMATION ABOUT SUSTAINML:
For any questions, please get in touch with
This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101070408.