Skip to content
/ ltt Public
forked from aangelopoulos/ltt

Learn then Test: Calibrating Predictive Algorithms to Achieve Risk Control

License

Notifications You must be signed in to change notification settings

clarafy/ltt

 
 

Repository files navigation

Learn then Test

Open In Colab

This repository will allow you to reproduce the experiments in the Learn then Test paper. For now, please e-mail me if you have trouble reproducing our results.

To install the dependencies for our experiments, run the following command from the root directory of the repository:

conda env create -f environment.yml
conda activate ltt

Then you should be able to run the experiment scripts.

The detectron code is different, and requires a separate set of dependencies. You will also need to make some modifications to the detectron source code.

In the /experiments/detectron folder, please execute

conda env create -f environment.yml
conda activate detectron2

Then you will need to edit the file

~/anaconda3/envs/detectron2/lib/python3.8/site-packages/detectron2/modeling/postprocessing.py

to add the following on line 67 (see the file experiments/detection/postprocessing.py):

results.roi_masks = roi_masks

Then the experiments should run.

About

Learn then Test: Calibrating Predictive Algorithms to Achieve Risk Control

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Jupyter Notebook 86.2%
  • Python 13.8%