search

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

Measuring the carbon footprint of downloads #6

Closed Grim-bot closed 2 years ago

Grim-bot commented 3 years ago

Context

When learning new ML skills or benchmarking ML algorithms, we often download a standard data set from TensorFlow and/or a pre-trained model from TensorFlow Hub. Presumably, the network traffic caused by these downloads leads to non-trivial greenhouse gas emissions.

Feature request

Thus, it would be useful to have a simple function for estimating the carbon footprint of downloads, like there are for computation and intra-server communication. It might also be helpful to rename the "communication costs" to make it clear that they are specifically for communication within a high-performance computing cluster.

Grim-bot commented 2 years ago

This issue was based on a misreading of the README