FLAPS

FLAPS (NeurIPS 2026)

This Hugging Face repository hosts checkpoints and prepared datasets for Flow Annealing Posterior Sampling for Function-Space Regression and Inverse Problems .

The source code and runnable scripts are available on GitHub:

https://github.com/yzshi5/FLAPS

Contents

PDE_inverse/checkpoints/
PDE_inverse/datasets/
Regression/checkpoints/FLAPS_prior/GP_gibbs_epoch_500.pt
Regression/checkpoints/FLAPS_prior/GP_matern_epoch_500.pt

The PDE inverse folders contain FNO surrogate checkpoints, FLAPS prior checkpoints, and small prepared test datasets used by the example scripts. The Regression folder currently includes the GP Gibbs and GP Matern prior checkpoints. For details, please visit the FLAPS GitHub repository.

• GitHub repository: https://github.com/yzshi5/FLAPS

Download Artifacts

Download these Hugging Face artifacts into the GitHub repo:

python - <<'PY'
from huggingface_hub import snapshot_download

snapshot_download(
    repo_id="Yaozhong/FLAPS",
    repo_type="model",
    local_dir=".",
    allow_patterns=[
        "config.json",
        "PDE_inverse/checkpoints/**",
        "PDE_inverse/datasets/**",
        "Regression/checkpoints/FLAPS_prior/GP_gibbs_epoch_500.pt",
        "Regression/checkpoints/FLAPS_prior/GP_matern_epoch_500.pt",
    ],
)
PY

Full PDE Dataset Download and Preprocessing

The helper Python files for downloading and preprocessing the full PDE datasets are under:

PDE_inverse/datasets/

Download the original PDE data shards:

cd PDE_inverse
python datasets/download_dataset.py all --output-dir datasets/PDE_inverse

Convert Hugging Face Arrow shards to .npy files:

python datasets/inital_process.py \
  --input-root datasets/PDE_inverse \
  --output-root datasets/PDE_inverse_npy

Citation

If you use this code or artifacts, please cite the associated FLAPS paper when available.

@article{shi2026flow,
  title={Flow Annealing Posterior Sampling for Function-Space Regression and Inverse Problems},
  author={Shi, Yaozhong and Ross, Zachary E and Yue, Yisong},
  journal={arXiv preprint arXiv:2606.22346v1},
  year={2026}
}