# PDEBench 1D Compressible Navier-Stokes Solutions

## Facts

- Provider: Makoto Takamoto et al. (NEC Labs Europe, University of Stuttgart)
- License: CC-BY-4.0 (https://darus.uni-stuttgart.de/api/datasets/:persistentId/?persistentId=doi:10.18419/darus-2986)
- Size: 7 files
- Format: HDF5 (in sampled files)
- Stored at: Hugging Face, https://huggingface.co/datasets/pdebench/1D-Compressible-Navier-Stokes
- Access: public, readable without an account
- DOI: https://doi.org/10.18419/darus-2986
- Category: Physics
- Source: https://arxiv.org/abs/2210.07182
- Page: https://scigantic.com/archive/d88f01f7-cd2b-53c9-b7ca-988af4b37367
- Keywords: Physics, CFD, PDE simulation, benchmark, HDF5, Hugging Face

## About

PDEBench (Takamoto, Praditia, Leiteritz, MacKinlay, Alesiani, Pfluger and Niepert; NEC Labs Europe and University of Stuttgart, NeurIPS 2022 Datasets and Benchmarks) is a benchmark of numerically simulated partial differential equations for scientific machine learning, used to compare neural operators such as FNO and U-Net and physics-informed networks. This repository holds the 1D compressible Navier-Stokes (CFD) training files, with density, velocity and pressure evolving under random initial conditions or shock-tube initial conditions, at different shear and bulk viscosities. Measured 2026-11-14, it contains five HDF5 files: random initial conditions with periodic boundaries at Eta=Zeta=0.01, 0.1 and 1e-8, random initial conditions with transmissive boundaries at 1e-8, and shock initial conditions with transmissive boundaries at 1e-8, 12 to 25 GB each and about 94 GB in total. Arrays follow the PDEBench convention [batch, time, x, variables] and open with h5py; the PDEBench GitHub repository provides loaders and baselines. The data are published on the University of Stuttgart DaRUS repository (doi:10.18419/darus-2986) under CC-BY-4.0.

## Load the data

```python
from huggingface_hub import hf_hub_download
path = hf_hub_download(repo_id="pdebench/1D-Compressible-Navier-Stokes", filename="1D_CFD_Rand_Eta0.01_Zeta0.01_periodic_Train.hdf5", repo_type="dataset")
print(path)
```

## Data structure

File types in sampled files: HDF5

Example file: `1D_CFD_Rand_Eta0.01_Zeta0.01_periodic_Train.hdf5`

## Open it

- In a hosted notebook with the data mounted: https://scigantic.com/archive/d88f01f7-cd2b-53c9-b7ca-988af4b37367
- Full record as JSON (schema card included): https://api.scigantic.com/api/archives/d88f01f7-cd2b-53c9-b7ca-988af4b37367
- Over MCP: `get_schema_card` with id `d88f01f7-cd2b-53c9-b7ca-988af4b37367` at https://api.scigantic.com/mcp

## Related topics

- [Benchmark datasets](https://scigantic.com/topics/benchmark)
- [HDF5 datasets](https://scigantic.com/topics/hdf5)
- [Physics datasets](https://scigantic.com/topics/physics)
- [CFD datasets](https://scigantic.com/topics/cfd)
