DomainBed is a PyTorch suite containing benchmark datasets and algorithms for domain generalization, as introduced in In Search of Lost Domain Generalization.

Full results for commit 7df6f06 in LaTeX format available here.

The currently available algorithms are:

• Empirical Risk Minimization (ERM, Vapnik, 1998)
• Invariant Risk Minimization (IRM, Arjovsky et al., 2019)
• Group Distributionally Robust Optimization (GroupDRO, Sagawa et al., 2020)
• Interdomain Mixup (Mixup, Yan et al., 2020)
• Marginal Transfer Learning (MTL, Blanchard et al., 2011-2020)
• Meta Learning Domain Generalization (MLDG, Li et al., 2017)
• Maximum Mean Discrepancy (MMD, Li et al., 2018)
• Deep CORAL (CORAL, Sun and Saenko, 2016)
• Domain Adversarial Neural Network (DANN, Ganin et al., 2015)
• Conditional Domain Adversarial Neural Network (CDANN, Li et al., 2018)
• Style Agnostic Networks (SagNet, Nam et al., 2020)
• Adaptive Risk Minimization (ARM, Zhang et al., 2020), contributed by @zhangmarvin
• Variance Risk Extrapolation (VREx, Krueger et al., 2020), contributed by @zdhNarsil
• Representation Self-Challenging (RSC, Huang et al., 2020), contributed by @SirRob1997
• Spectral Decoupling (SD, Pezeshki et al., 2020)
• Learning Explanations that are Hard to Vary (AND-Mask, Parascandolo et al., 2020)
• Out-of-Distribution Generalization with Maximal Invariant Predictor (IGA, Koyama et al., 2020)
• Gradient Matching for Domain Generalization (Fish, Shi et al., 2021)
• Self-supervised Contrastive Regularization (SelfReg, Kim et al., 2021)
• Smoothed-AND mask (SAND-mask, Shahtalebi et al., 2021)
• Invariant Gradient Variances for Out-of-distribution Generalization (Fishr, Rame et al., 2021)
• Learning Representations that Support Robust Transfer of Predictors (TRM, Xu et al., 2021)
• Invariance Principle Meets Information Bottleneck for Out-of-Distribution Generalization (IB-ERM , Ahuja et al., 2021)
• Invariance Principle Meets Information Bottleneck for Out-of-Distribution Generalization (IB-IRM, Ahuja et al., 2021)
• Optimal Representations for Covariate Shift (CAD & CondCAD, Ruan et al., 2022), contributed by @ryoungj
• Quantifying and Improving Transferability in Domain Generalization (Transfer, Zhang et al., 2021), contributed by @Gordon-Guojun-Zhang
• Invariant Causal Mechanisms through Distribution Matching (CausIRL with CORAL or MMD, Chevalley et al., 2022), contributed by @MathieuChevalley

Send us a PR to add your algorithm! Our implementations use ResNet50 / ResNet18 networks (He et al., 2015) and the hyper-parameter grids described here.

The currently available datasets are:

• RotatedMNIST (Ghifary et al., 2015)
• ColoredMNIST (Arjovsky et al., 2019)
• VLCS (Fang et al., 2013)
• PACS (Li et al., 2017)
• Office-Home (Venkateswara et al., 2017)
• A TerraIncognita (Beery et al., 2018) subset
• DomainNet (Peng et al., 2019)
• A SVIRO (Dias Da Cruz et al., 2020) subset
• WILDS (Koh et al., 2020) FMoW (Christie et al., 2018) about satellite images
• WILDS (Koh et al., 2020) Camelyon17 (Bandi et al., 2019) about tumor detection in tissues

Send us a PR to add your dataset! Any custom image dataset with folder structure dataset/domain/class/image.xyz is readily usable. While we include some datasets from the WILDS project, please use their official code if you wish to participate in their leaderboard.

Model selection criteria differ in what data is used to choose the best hyper-parameters for a given model:

• IIDAccuracySelectionMethod: A random subset from the data of the training domains.
• LeaveOneOutSelectionMethod: A random subset from the data of a held-out (not training, not testing) domain.
• OracleSelectionMethod: A random subset from the data of the test domain.

Download the datasets:

python3 -m domainbed.scripts.download \
       --data_dir=./domainbed/data

Train a model:

python3 -m domainbed.scripts.train\
       --data_dir=./domainbed/data/MNIST/\
       --algorithm IGA\
       --dataset ColoredMNIST\
       --test_env 2

Launch a sweep:

python -m domainbed.scripts.sweep launch\
       --data_dir=/my/datasets/path\
       --output_dir=/my/sweep/output/path\
       --command_launcher MyLauncher

Here, MyLauncher is your cluster's command launcher, as implemented in command_launchers.py. At the time of writing, the entire sweep trains tens of thousands of models (all algorithms x all datasets x 3 independent trials x 20 random hyper-parameter choices). You can pass arguments to make the sweep smaller:

python -m domainbed.scripts.sweep launch\
       --data_dir=/my/datasets/path\
       --output_dir=/my/sweep/output/path\
       --command_launcher MyLauncher\
       --algorithms ERM DANN\
       --datasets RotatedMNIST VLCS\
       --n_hparams 5\
       --n_trials 1

After all jobs have either succeeded or failed, you can delete the data from failed jobs with python -m domainbed.scripts.sweep delete_incomplete and then re-launch them by running python -m domainbed.scripts.sweep launch again. Specify the same command-line arguments in all calls to sweep as you did the first time; this is how the sweep script knows which jobs were launched originally.

To view the results of your sweep:

python -m domainbed.scripts.collect_results\
       --input_dir=/my/sweep/output/path

DomainBed includes some unit tests and end-to-end tests. While not exhaustive, but they are a good sanity-check. To run the tests:

python -m unittest discover

By default, this only runs tests which don't depend on a dataset directory. To run those tests as well:

DATA_DIR=/my/datasets/path python -m unittest discover

This source code is released under the MIT license, included here.