File size: 3,204 Bytes
d8cc85c
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
20ee1e1
d8cc85c
 
20ee1e1
 
d8cc85c
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
---
license: other
#User-Defined Tags
tags:
  - Operator Learning
  - Neural PDE Solvers
language:
  - en
  - zh
---
<p align="center">
  <strong>
    <span style="font-size: 30px;">PDENNEval</span>
  </strong>
</p>

## Dataset Description

PDENNEval is a comprehensive dataset for evaluating neural-network-based PDE solving methods, introduced in an IJCAI 2024 paper. It covers function learning and operator learning tasks and includes 15 types of PDE problems across multiple scientific domains, including fluids, materials, finance, and electromagnetics.

The dataset consists of 10 PDEBench data files and 6 self-generated data files, totaling approximately 286.9 GB. It can be used for model sanity checks, training, evaluation, and comparisons across problems.

Paper: [PDENNEval: A Comprehensive Evaluation of Neural Network Methods for Solving PDEs](https://www.ijcai.org/proceedings/2024/573)

## Supported Tasks

| Scenario | Description |
|---|---|
| PDE solution field prediction | Predict PDE solutions from equation parameters, coordinates, or initial conditions. |
| Function learning evaluation | Compare neural networks' ability to solve individual PDE problems. |
| Operator learning evaluation | Evaluate model generalization across different input conditions and equation tasks. |
| Cross-domain scientific computing | Cover fluid, diffusion-reaction, materials, finance, and electromagnetics problems. |

## Dataset Format and Structure

All data files are stored in HDF5 format with the `.hdf5` or `.h5` extension:

```text
data/
  PDEBench data files
  Self-generated PDE data files
```

The data includes coordinates, time, coefficient fields, boundary conditions, and solution fields for one-, two-, and three-dimensional PDEs. It covers problems such as Advection, Burgers, Diffusion-Reaction, Darcy Flow, Shallow Water, Allen-Cahn, Cahn-Hilliard, Navier-Stokes, Euler, Maxwell, and Black-Scholes-Barenblatt.

Fields and tensor shapes vary across PDEs. Refer to the data files and `metadata/pdenneval_schema.json` for the specific schema.

## How to Use the Dataset

This dataset is designed for the `OneScience-Group/PDENNEval` model. Download the dataset and model:

```bash
hf download --dataset OneScience-Group/pdenneval --local-dir ./pdenneval
hf download --model OneScience-Group/PDENNEval --local-dir ./PDENNEval
```

## 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

- Original PDENNEval paper: [PDENNEval: A Comprehensive Evaluation of Neural Network Methods for Solving PDEs](https://doi.org/10.24963/ijcai.2024/573)
- Original PDEBench paper: [PDEBench: An Extensive Benchmark for Scientific Machine Learning](https://arxiv.org/abs/2210.07182)
- PDEBench dataset: [PDEBench Datasets](https://doi.org/10.18419/darus-2986)
- This repository retains source attribution. Before public distribution or republication, confirm the licensing requirements with the upstream projects.