https://www.whitehouse.gov/ostp/news-updates/2022/09/08/fact-sheet-climate-and-energy-implications-of-crypto-assets-in-the-united-states/ 

The United States is estimated to host about a third of global crypto-asset operations, which currently consume about 0.9% to 1.7% of total U.S. electricity usage. This range of electricity usage is similar to all home computers or residential lighting in the United States.

Global electricity generation for the crypto-assets with the largest market capitalizations resulted in a combined 140 ± 30 million metric tons of carbon dioxide per year (Mt CO₂/y), or about 0.3% of global annual greenhouse gas emissions. Crypto-asset activity in the United States is estimated to result in approximately 25 to 50 Mt CO₂/y, which is 0.4% to 0.8% of total U.S. greenhouse gas emissions. This range of emissions is similar to emissions from diesel fuel used in railroads in the United States.

codecarbon 2.3.2

 https://pypi.org/project/codecarbon/

https://mlco2.github.io/codecarbon/methodology.html# 

Carbontracker: Tracking and Predicting the Carbon Footprint of Training Deep Learning Models

 https://arxiv.org/abs/2007.03051

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

 https://arxiv.org/abs/2002.05651 

Quantifying the Carbon Emissions of Machine Learning

 https://arxiv.org/abs/1910.09700