EPFLiGHT / cumulator

A tool to quantify and report the carbon footprint of machine learning computations and communication
MIT License
21 stars 5 forks source link

Regularly (yearly ?) update the value used as fall-back / default for the carbon intensity of electric power. #19

Open Grim-bot opened 2 years ago

Grim-bot commented 2 years ago

As time goes on, carbon intensity of electricity from 2014 will hopefully be more and more outdated. Can this value be automatically fetched from some public source and updated regularly, e.g., every one or two years?

vincenzopecorella commented 2 years ago

Cumulator bases its computation on the monthly updated https://github.com/owid/energy-data?country=. We already included a script to modify the dataset obtained from the source above to be directly used with Cumulator. I've just updated the README to make it explicit.

Grim-bot commented 2 years ago

Thank you! I was referring to this part of the readme:

Cumulator will try to set the carbon intensity value based on the geographical position of the user. In case the detection fails the default value will be set. It is possible to manually modify the default value.

self.carbon_intensity = 447 <- conversion to carbon footprint: average carbon intensity value in gCO2eq/kWh in the EU in 2014

Is this fallback (default) value also updated using https://github.com/owid/energy-data ?

Grim-bot commented 2 years ago

@vincenzopecorella

Then we can close this issue

vincenzopecorella commented 2 years ago

Hey! We didn't modify the default value, it is the same computed by @tristantreb. About hardware specs you are right, we took them from techpowerup.com and can be worth mentioning in the README. We can't however modify this repo since we don't have the rights here.

By the way, I just noticed that on https://github.com/owid/energy-data they have updated the dataset, it can be nice to swap old dataset with the updated one. We made some functions to ease this process and you can find them under /src/cumulator/countries_data/country_dataset_helpers.py