---
license: other
#User-Defined Tags
tags:
  - Multi-Physics PDE Benchmark
  - Surrogate Modeling
language:
  - en
  - zh
---
<p align="center">
  <strong>
    <span style="font-size: 30px;">CFD Benchmark</span>
  </strong>
</p>

## Dataset Description

CFD Benchmark is a comprehensive benchmark dataset for training and evaluating neural PDE solvers. It contains six standard tasks on regular grids, structured grids, and irregular geometries. The dataset was used for unified evaluation in the ICML 2024 paper *Transolver*, and some of its data originates from the FNO and Geo-FNO works.

The dataset contains six subsets: `airfoil`, `darcy`, `elasticity`, `ns`, `pipe`, and `plas`.

Paper: [Transolver: A Fast Transformer Solver for PDEs on General Geometries](https://proceedings.mlr.press/v235/wu24r.html)

## Supported Tasks

| Subset | Task |
|---|---|
| `airfoil` | Steady-state airfoil flow-field prediction. |
| `darcy` | Darcy equation solution-field prediction. |
| `elasticity` | Stress prediction for random-cell structures. |
| `ns` | Navier-Stokes flow-field time-series prediction. |
| `pipe` | Steady-state pipe flow-field prediction. |
| `plas` | Plasticity dynamics response prediction. |

## Dataset Format and Structure

The data is located in the `data/` directory and organized by task:

```text
data/
  airfoil/
  darcy/
  elasticity/
  ns/
  pipe/
  plas/
```

The data files use NumPy NPY and MATLAB v5 formats. The Airfoil, Pipe, and Elasticity subsets contain grid coordinates, geometry, or flow-field arrays; the Darcy and Navier-Stokes subsets use MATLAB files to store coefficient and solution fields; and the Plasticity subset contains load inputs and response outputs over multiple time steps.

See `metadata/schema.yaml` for the shape and dtype of each file and the MATLAB variable structures.

## How to Use the Dataset

This dataset has been adapted for the `OneScience-Group/CFD_Benchmark` model. Download the dataset and model:

```bash
hf download --dataset OneScience-Group/cfd_benchmark --local-dir ./cfd_benchmark
hf download --model OneScience-Group/CFD_Benchmark --local-dir ./CFD_Benchmark
```

Quickly check file presence, sizes, and readability:

```bash
cd cfd_benchmark
python scripts/validate_dataset.py --skip-sha256
```

Remove `--skip-sha256` to perform a complete SHA256 validation.

## Official OneScience Information

| Platform | OneScience Main Repository | Skills Repository |
|---|---|---|
| Gitee | https://gitee.com/onescience-ai/onescience | https://gitee.com/onescience-ai/oneskills |
| GitHub | https://github.com/onescience-ai/OneScience | https://github.com/onescience-ai/oneskills |

## Citation and License

- Six standard tasks: [Transolver](https://proceedings.mlr.press/v235/wu24r.html)
- Darcy and Navier-Stokes data source: [Fourier Neural Operator](https://openreview.net/forum?id=c8P9NQVtmnO)
- Elasticity, Plasticity, Airfoil, and Pipe data source: [Geo-FNO](https://jmlr.org/papers/v24/23-0064.html)
- This repository retains the data-source information. Before public distribution or redistribution, verify the licensing requirements with the upstream projects.