CFD_Benchmark
Model Overview
CFD_Benchmark is an open-source deep-learning benchmark library for research on neural partial differential equation (PDE) solvers. It extends Tsinghua University's open-source Neural-Solver-Library with Distributed Data Parallel (DDP) training support, additional models, and new datasets, while retaining the original neural-operator and physical-field modeling framework. The library supports neural PDE solver evaluation, deep-learning research for CFD, multi-model performance comparisons, large-scale distributed training experiments, physical simulation dataset development, and algorithm benchmarking.
The library currently supports the following benchmarks:
Six standard benchmarks from [FNO] and [geo-FNO]
PDEBench [NeurIPS 2022 Track dataset and benchmark] for autoregressive tasks
The ShapeNet-Car dataset [TOG 2018] for industrial design benchmarks
The BubbleML [Multiphase Multiphysics Dataset] for studying multiphysics phase-transition phenomena
Supported Neural Solvers
The following neural PDE solvers are supported:
Transolver - Transolver: A Fast Transformer Solver for PDEs on General Geometries
[ICML 2024] [Code]
ONO - Improved Operator Learning by Orthogonal Attention
[ICML 2024] [Code]
Factformer - Scalable Transformer for PDE Surrogate Modeling
[NeurIPS 2023] [Code]
U-NO - U-NO: U-shaped Neural Operators
[TMLR 2023] [Code]
LSM - Solving High-Dimensional PDEs with Latent Spectral Models
[ICML 2023] [Code]
GNOT - GNOT: A General Neural Operator Transformer for Operator Learning
[ICML 2023] [Code]
F-FNO - Factorized Fourier Neural Operators
[ICLR 2023] [Code]
U-FNO - An enhanced Fourier neural operator-based deep-learning model for multiphase flow
[Advances in Water Resources 2022] [Code]
Galerkin Transformer - Choose a Transformer: Fourier or Galerkin
[NeurIPS 2021] [Code]
MWT - Multiwavelet-based Operator Learning for Differential Equations
[NeurIPS 2021] [Code]
FNO - Fourier Neural Operator for Parametric Partial Differential Equations
[ICLR 2021] [Code]
Transformer - Attention Is All You Need
[NeurIPS 2017] [Code]
GFNO - Group Equivariant Fourier Neural Operators for Partial Differential Equations
[2023 Poster] [Code]
Several vision architectures also serve as effective baselines for structured-geometry tasks:
Swin Transformer - Swin Transformer: Hierarchical Vision Transformer using Shifted Windows [ICCV 2021] [Code] U-Net - U-Net: Convolutional Networks for Biomedical Image Segmentation [MICCAI 2015] [Code]
Several established geometric deep-learning models are included for design tasks:
Graph-UNet - Graph U-Nets [ICML 2019] [Code] GraphSAGE - Inductive Representation Learning on Large Graphs [NeurIPS 2017] [Code] PointNet - PointNet: Deep Learning on Point Sets for 3D Classification and Segmentation [CVPR 2017] [Code]
The library also includes the following graph neural network:
MeshGraphNet LEARNING MESH-BASED SIMULATION WITH GRAPH NETWORKSICLR 2021 [Code]
Use Cases
Use Case Description Neural PDE solver evaluation Train, run inference with, and compare models such as FNO, Transolver, GNOT, ONO, and U-NO through a unified workflow Autoregressive physical prediction Predict the temporal evolution of PDE states step by step using datasets such as PDEBench Multiphysics modeling Study multiphase flows, multiphysics coupling, and phase-transition phenomena using datasets such as BubbleML ModelScope/OneCode execution Download the standalone model package, install its dependencies, and run the provided scripts
Usage
1. OneCode
Use the online OneCode environment for an intelligent, one-click AI for Science (AI4S) programming experience:
2. Manual Setup
Hardware Requirements
A GPU or DCU is recommended.
A CPU can be used for import checks and small-scale pipeline validation, but full training and inference will be slow.
DCU users must install DTK in advance. DTK 25.04.2 or later, or the OneScience-recommended version for the target cluster, is recommended.
Download the Model Package
1 hf download OneScience-Group/CFD_Benchmark --local-dir ./CFD_Benchmark
2 cd CFD_Benchmark
Set Up the Runtime Environment
DCU Environment
1 # Activate DTK and Conda first
2 conda create -n onescience311 python = 3.11 -y
3 conda activate onescience311
4 # Installation with uv is also supported
5 pip install onescience [ cfd-dcu ] -i http://mirrors.onescience.ai:3141/pypi/simple/ --trusted-host mirrors.onescience.ai
GPU Environment
1 # Activate Conda first
2 conda create -n onescience311 python = 3.11 -y libstdcxx-ng = 12 libgcc-ng = 12 gcc_linux-64 = 12 gxx_linux-64 = 12
3 conda activate onescience311
4 # Installation with uv is also supported
5 pip install onescience [ cfd-gpu ] -i http://mirrors.onescience.ai:3141/pypi/simple/ --trusted-host mirrors.onescience.ai
Training Data
Use the dataset links in the benchmark overview above to download the required data.
The six standard benchmark datasets from
[FNO] and
[geo-FNO] are available from
this link .
The PDEBench
[NeurIPS 2022 Track dataset and benchmark] , used for benchmarking autoregressive tasks, is available from
this link .
The ShapeNet-Car
[TOG 2018] benchmark dataset for industrial design tasks is available from
[this link] .
The BubbleML
[Multiphase Multiphysics Dataset] , designed for research on multiphysics phase-transition phenomena, is available from
[this link] .
The OneScience community also provides the cfd_benchmark dataset for training. Download it with the command below and verify that the data path in conf/config.yaml is configured correctly:
hf download --repo-type dataset OneScience-Group/cfd_benchmark --local-dir ./dataTraining
Model Weights
This repository will provide weights trained on the OneScience cfd_benchmark dataset in the weights/ directory. The weights will be uploaded soon.
Inference
python scripts/inference.pyInference loads the trained weights referenced by paths.weight_path and writes the metrics to:
./results/{train.save_name}/metrics.jsonOfficial OneScience Resources
Citations and License
Reference repository: Neural-Solver-Library .
This repository retains source attribution and has been adapted for automated execution through OneScience and ModelScope.
