Skip to content

apourchot/to_share_or_not_to_share

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

12 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

To Share or Not To Share: A Comprehensive Appraisal of Weight-Sharing

This directory contains all the code necessary to reproduce the results of the paper: https://arxiv.org/abs/2002.04289v1

Main Dependencies:

Steps to reproduce the results:

  1. Download the nasbench dataset and put under the "datasets" folder
  2. Run all the cells of the notebook "datasets.ipynb": this will create all the different search spaces
  3. Run individual trainings and evaluations with python main.py --search_space SEARCH_SPACE after having set parameters as wanted in "main.py". Seeds used to generate the results of the paper can be found in the "figures.ipynb" notebook.
  4. generate the different figures using the "figures.ipnyb" notebook, after having arranged the files resulting from the evaluations as in the "results_paper" directory

Visualization

  • Examples of architectures coming from each dataset can be generated at the end of the "datasets.ipynb" notebook.

  • All the curves of the paper (and more) can be generated and observed in the "figures.ipnyb" notebook.

About

No description, website, or topics provided.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published