This repository holds the Pytorch implementation of Ensemble Pruning for Out-of-distribution Generalization by Fengchun Qiao and Xi Peng. If you find our code useful in your research, please consider citing:
@inproceedings{qiao2024tep,
title={Ensemble Pruning for Out-of-distribution Generalization},
author={Fengchun Qiao and Xi Peng},
booktitle={International Conference on Machine Learning (ICML)},
year={2024}
}
Our code is adapted from the open-source DomainBed github and DiWA github
- python == 3.7.10
- torch == 1.8.1
- torchvision == 0.9.1
- numpy == 1.20.2
Please follow DiWA github to obtain pre-trained individual models
We average the weights selected by our method
python -m domainbed.scripts.diwa\
--data_dir=/my/data/dir/\
--output_dir=/my/sweep/output/path\
--dataset TerraIncognita\
--test_env ${test_env}\
--weight_selection TEP\
--trial_seed ${trial_seed}Please contact Fengchun Qiao (fengchun@udel.edu) for any question.