search

mlco2 / codecarbon

Track emissions from Compute and recommend ways to reduce their impact on the environment.
https://mlco2.github.io/codecarbon
MIT License
1.12k stars 173 forks source link

Measurement of Energy Consumption during Model Training in Multi-User and Cloud Environments #411

Open RafiullahOmar opened 1 year ago

RafiullahOmar commented 1 year ago

Description Thank you for the great package. I have a question regarding the scenario where a server is used by multiple users and one user trains models. Is it possible to accurately measure the energy consumed during model training in such a scenario? Additionally, what if we train a model on Colab or on cloud? Can we accurately measure the energy consumption in those cases as well? Thank you

benoit-cty commented 1 year ago

Hello, if the hardware is shared across users, yes the electric consumption of other users will be use for your emissions.

That's why we are working on https://github.com/mlco2/codecarbon/pull/316 that allow to set tracking_mode to process and monitor only the emissions of your process.

You can test it, it should work.