-
Eco2AI: carbon emissions tracking of machine learning models as the first step towards sustainable AI
Abstract: The size and complexity of deep neural networks continue to grow exponentially, significantly increasing energy consumption for training and inference by these models. We introduce an open-source package eco2AI to help data scientists and researchers to track energy consumption and equivalent CO2 emissions of their models in a straightforward way. In eco2AI we put emphasis on accuracy of energy co… ▽ More
Submitted 3 August, 2022; v1 submitted 31 July, 2022; originally announced August 2022.
Comments: Source code for eco2AI package (energy consumption and carbon emission tracker of code in python) is available at: https://github.com/sb-ai-lab/Eco2AI , the package is also available at PyPi: https://pypi.org/project/eco2ai/