Packed-Ensembles Experiments

Library of model configurations to reproduce the Packed-Ensembles paper.

These model configurations will work until at least torch-uncertainty==0.10.1.

Usage examples

Classification

Examples:

• Training a standard ResNet18 model as in Packed-Ensembles for Efficient Uncertainty Estimation:

python main.py fit --config configs/resnet18/standard.yaml

• Training Packed-Ensembles ResNet50 model as in Packed-Ensembles for Efficient Uncertainty Estimation:

python main.py fit --config configs/resnet50/packed.yaml

Regression

Example:

cd regression/uci_datasets
python main.py fit --config configs/boston/mlp/packed_ensembles.yaml

Contact us if you need more configuration files (for instance for the rest of UCI benchmark). Also look at the organization's repositories for more experiment configuration files.

Citation

If you find this repository useful for your research, please consider citing

@article{laurent2022packed,
  title={Packed-ensembles for efficient uncertainty estimation},
  author={Laurent, Olivier and Lafage, Adrien and Tartaglione, Enzo and Daniel, Geoffrey and Martinez, Jean-Marc and Bursuc, Andrei and Franchi, Gianni},
  journal={ICLR},
  year={2023}
}
@inproceedings{lafage2025torch,
  title={Torch-Uncertainty: Deep Learning Uncertainty Quantification},
  author={Lafage, Adrien and Laurent, Olivier and Gabetni, Firas and Franchi, Gianni},
  booktitle={NeurIPS D&B}
  year={2025}
}