GreenFlux.jl

Functions that allow to monitor the power draw from CPU, GPU's and DDR3-RAM in kWh and approximate the number of non-embedding Floating Point operations that a Flux model will make.

Go to file

Eduardo Cueto Mendoza 46aef70cb1 Solved problem with RAM calculations		2020-07-28 16:07:21 -06:00
.github/workflows	Failing in x86 removed temporally	2020-06-30 23:19:07 -06:00
docs	CompatHelper: bump compat for "Flux" to "0.11"	2020-07-11 00:34:08 +00:00
src	Solved problem with RAM calculations	2020-07-28 16:07:21 -06:00
test	Solved a issue with linux, still have to check linux if nvidia-smi is, powerstat of free are missing	2020-07-05 16:56:31 -06:00
.gitignore	Added new documentation folders for v0.0.4	2020-06-27 10:46:44 -06:00
.travis.yml	Removed Coveralls from TravisCI	2020-06-27 11:17:56 -06:00
LICENSE	Files generated by PkgTemplates	2020-06-27 09:59:08 -06:00
Manifest.toml	CompatHelper: bump compat for "Flux" to "0.11"	2020-07-11 00:34:08 +00:00
Project.toml	Solved issue with W suffix in GPU powerdraw	2020-07-28 11:59:11 -06:00
README.md	Added a only working in Linux advisory	2020-07-09 19:25:38 -06:00
green-flux-logo.png	Updated read me with a description and added a new Logo	2020-07-05 19:09:12 -06:00

README.md

GreenFlux adds Green AI functionalities to Flux so you can make more precise reports on your peer reviewed papers or conference submissions of the electricity consumed by your model during training and number of non-embedding Floating Point OPerations that your Flux model will require during forward and back propagation.

To install just do:

] add GreenFlux

Due to the limited hardware and driver accessibility this framework has been tested and works currently in Linux.

Tested Successfully On

GPUs:

- NVIDIA GeForce GTX 1050

CPUs:

- Intel(R) Core(TM) i7-7700HQ CPU @ 2.80GHz

OS:

- Ubuntu 18.04.1