bouthilx / ml-method-video

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Videos for the paper Accounting for Variance in Machine Learning Benchmarks @ MLSys2021

This repository contains the code to create the short and oral videos of our paper at MLSys2021.

The animation are all generated with matplotlib.

It does not generate the audio-track unfortunately, this part was recorded. :)

Install

For installation, simply install the requirements with pip.

$ pip install -r requirements.txt

The data used for some of the visualizations is available on dropbox at https://www.dropbox.com/sh/cwcjd91gw9wu6wv/AABg8h4Iq8ZQUuhoIkpDNNW-a?dl=0.

The default path used by the script is ~/Dropbox/Olympus-Data. When running the script, make sure to pass --data-folder with correct path if you save the data elsewhere.

Execution

For the short video

$ python procedure.py --fps 60 --dpi 100 --parallel --concat

For the oral video

$ python mlsys2021.py --fps 60 --dpi 100 --parallel --concat

About


Languages

Language:Python 100.0%