diff --git a/README.en.md b/README.en.md new file mode 100644 index 0000000000000000000000000000000000000000..48b4b391787547b7a8cf9f24f88ae6d1a87d9af1 --- /dev/null +++ b/README.en.md @@ -0,0 +1,121 @@ +# OneScience Examples + +This project is a collection of AI for Science (AI4S) agent generation examples on the OneScience platform, covering multiple scientific domains including biosciences, computational fluid dynamics, earth and meteorological prediction, materials chemistry, and structural engineering. + +## Project Domains + +| Domain | Example | Description | +|--------|---------|-------------| +| **Biosciences** | alphafold | Protein structure prediction | +| **Computational Fluid Dynamics (CFD)** | DeepCFD, Transolver-RAE2822 | Fluid simulation and airfoil design | +| **Earth and Meteorology** | FuXi, Pangu, Pangu-FourCastNet-Hybrid | Weather forecasting and global prediction | +| **Materials Chemistry** | mace | Molecular structure and property prediction | +| **Structural Engineering** | DEM_for_plasticity | Granular material plasticity simulation | + +## Directory Structure + +``` +onescience-examples/ +├── biosciences/ +│ └── alphafold/ # Protein structure prediction +├── cfd/ +│ ├── DeepCFD/ # Deep learning-based fluid simulation +│ └── Transolver-RAE2822/ # Transolver airfoil dataset +├── earth/ +│ ├── fuxi/ # FuXi weather forecasting system +│ │ └── project/ +│ │ ├── config.py # Configuration management +│ │ ├── model.py # FuXi model definition +│ │ ├── train.py # Training pipeline +│ │ └── eval.py # Evaluation metrics +│ ├── pangu/ # Meteorological forecasting model +│ └── pangu-fourcastnet-hybrid/ # Hybrid forecasting model +├── MaterialsChemistry/ +│ └── mace/ # Molecular structure prediction +└── structural/ + └── DEM_for_plasticity/ # Discrete element analysis +``` + +## System Requirements + +- Python 3.8+ +- PyTorch 1.10+ +- CUDA 11.0+ (GPU support) + +## Installation + +```bash +# Clone the repository +git clone https://gitee.com/onescience-ai/onescience-examples.git +cd onescience-examples + +# Install base dependencies +pip install torch numpy scipy +``` + +## Detailed Examples by Domain + +### Computational Fluid Dynamics (CFD) + +#### Transolver-RAE2822 + +An adapted example of the RAE2822 airfoil dataset using the Transolver model, including a complete dataset, training, and inference pipeline. + +**Core Modules:** +- `RAE2822.py` - Dataset class (`RAE2822Dataset`) and data pipeline (`RAE2822Datapipe`) +- `train.py` - Training entry point +- `inference.py` - Inference entry point +- `conf/transolver_rae2822.yaml` - Training configuration file + +**Usage:** +```bash +cd cfd/Transolver-RAE2822 +python train.py --config conf/transolver_rae2822.yaml +python inference.py --checkpoint model.pt +``` + +### Earth and Meteorological Prediction (Earth) + +#### FuXi Cascading Forecast System + +FuXi is a cascading machine learning system for weather forecasting. + +**Core Modules:** +- `project/config.py` - `FuXiConfig` configuration class +- `project/model.py` - `FuXi` model definition (`nn.Module`) +- `project/train.py` - `WeatherDataset` dataset class + training function +- `project/eval.py` - Evaluation metrics (RMSE, ACC, etc.) + +#### Pangu-FourCastNet Hybrid Model + +A hybrid forecasting architecture combining Pangu and FourCastNet: +- `HybridPanguFourCastNetDay7` - Hybrid model class +- Supports 7-day forecast lead time +- Pangu default resolution: 13 vertical levels, 721×1440 for upper atmosphere, 721×1440 for surface + +### Materials Chemistry (MaterialsChemistry) + +#### MACE + +MACE (Machine Learning for Atomic Cluster Expansion) is a machine learning model for computing molecular potential energy surfaces. + +### Biosciences (Biosciences) + +#### AlphaFold + +A deep learning model based on the Transformer architecture for protein structure prediction. + +### Structural Engineering (Structural) + +#### DEM_for_plasticity + +Discrete Element Method (DEM) for simulating plasticity in granular materials. + +## License + +This project follows the terms specified in the LICENSE file. + +## References + +- OneScience Framework: https://gitee.com/onescience-ai +- Refer to the README.md files of each subproject for related papers \ No newline at end of file diff --git a/README.md b/README.md new file mode 100644 index 0000000000000000000000000000000000000000..4b1f9608b97bfb7c3e72932ec81f17f0b0b44dfc --- /dev/null +++ b/README.md @@ -0,0 +1,123 @@ + + +# OneScience Examples + +本项目是 OneScience 平台的 AI for Science (AI4S) 智能体生成案例集，涵盖生物科学、流体力学、地球气象预测、材料化学和结构工程等多个科学领域。 + +## 项目领域 + +| 领域 | 示例 | 说明 | +|------|------|------| +| **生物科学** | alphafold | 蛋白质结构预测 | +| **计算流体力学** | DeepCFD, Transolver-RAE2822 | 流体仿真与翼型设计 | +| **地球气象** | FuXi, Pangu, Pangu-FourCastNet-Hybrid | 气象预报与全球预测 | +| **材料化学** | mace | 分子结构与属性预测 | +| **结构工程** | DEM_for_plasticity | 颗粒材料塑性模拟 | + +## 目录结构 + +``` +onescience-examples/ +├── biosciences/ +│ └── alphafold/ # 蛋白质结构预测 +├── cfd/ +│ ├── DeepCFD/ # 深度学习流体仿真 +│ └── Transolver-RAE2822/ # Transolver 翼型数据集 +├── earth/ +│ ├── fuxi/ # FuXi 气象预报系统 +│ │ └── project/ +│ │ ├── config.py # 配置管理 +│ │ ├── model.py # FuXi 模型定义 +│ │ ├── train.py # 训练流程 +│ │ └── eval.py # 评估指标 +│ ├── pangu/ # 气象预报模型 +│ └── pangu-fourcastnet-hybrid/ # 混合预报模型 +├── MaterialsChemistry/ +│ └── mace/ # 分子结构预测 +└── structural/ + └── DEM_for_plasticity/ # 离散元分析 +``` + +## 环境要求 + +- Python 3.8+ +- PyTorch 1.10+ +- CUDA 11.0+ (GPU 支持) + +## 安装 + +```bash +# 克隆项目 +git clone https://gitee.com/onescience-ai/onescience-examples.git +cd onescience-examples + +# 安装基础依赖 +pip install torch numpy scipy +``` + +## 各领域示例详情 + +### 计算流体力学 (CFD) + +#### Transolver-RAE2822 + +基于 Transolver 模型的 RAE2822 翼型数据集适配示例，包含完整的数据集、训练和推理流程。 + +**核心模块：** +- `RAE2822.py` - 数据集类 (`RAE2822Dataset`) 和数据管道 (`RAE2822Datapipe`) +- `train.py` - 训练入口 +- ` inference.py` - 推理入口 +- `conf/transolver_rae2822.yaml` - 训练配置文件 + +**使用方式：** +```bash +cd cfd/Transolver-RAE2822 +python train.py --config conf/transolver_rae2822.yaml +python inference.py --checkpoint model.pt +``` + +### 地球气象预测 (Earth) + +#### FuXi 级联预报系统 + +FuXi 是一个级联机器学习气象预报系统。 + +**核心模块：** +- `project/config.py` - `FuXiConfig` 配置类 +- `project/model.py` - `FuXi` 模型定义 (`nn.Module`) +- `project/train.py` - `WeatherDataset` 数据集类 + 训练函数 +- `project/eval.py` - RMSE、ACC 等评估指标计算 + +#### Pangu-FourCastNet 混合模型 + +结合 Pangu 和 FourCastNet 的混合预报架构： +- `HybridPanguFourCastNetDay7` - 混合模型类 +- 支持 7 天预报 lead time +- Pangu 默认分辨率：上空 13层、721×1440，地表 721×1440 + +### 材料化学 (MaterialsChemistry) + +#### MACE + +MACE (Machine Learning for Atomic Cluster Expansion) 用于分子势能面计算的机器学习模型。 + +### 生物科学 (Biosciences) + +#### AlphaFold + +基于 Transformer 架构的蛋白质结构预测深度学习模型。 + +### 结构工程 (Structural) + +#### DEM_for_plasticity + +离散元方法 (DEM) 用于颗粒材料塑性模拟。 + +## 许可证 + +本项目遵循 LICENSE 文件中指定的条款。 + +## 参考资源 + +- OneScience 框架：https://gitee.com/onescience-ai +- 相关论文请查阅各子项目的 README.md 文件 \ No newline at end of file