Skip to content

rhubarbwu/ENAS-Experiments

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

56 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

ENAS Experiments

This codebase has been adapted to run experiments on improving/evaluating the robustness of the ENAS algorithm. Thanks to Melody Guan for the basis of work.

Requirements

pip install -r requirements.txt

Setup

Prepare the experiments in experiments/. You can name <experiment-set-name> whatever you like. We provide the following experiment sets for <experiment-set-url>.

Running

Include the set, experiment, and number of epochs as arguments. They will default to the baseline, space_0, and 300.

# experiment from baseline set
python driver.py baseline space_0 300

# experiment from poisoning set
python driver.py poisoning poisoning_0 300

# experiment from typed set
python driver.py neuralarts types_0 300

Checkpoints & Results

Results are CSV files stored under results/, also named after the set and experiment. Model checkpoints will be stored by a filename named after the set and experiment in checkpoints/.

  • Do not run the same experiment in two instances at once as their models will overwrite each other. Running different experiments from the same set is okay.
  • Experiments that conclude (i.e. epoch counter reaches num_epoch) will save their CSV with a timestamp in the filename at time of completion so they will not be overwritten.
  • If you're repeating an experiment that concluded before, delete the previous checkpoint.

About

ENAS codebase+experiments for AutoMLG@UTMIST.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Contributors 4

  •  
  •  
  •  
  •