[WIP] Add convert data script and docs

Author: BeingGod

docs/zh/examples/pangu_weather.md

@@ -0,0 +1,105 @@
+# Pangu-Weather
+
+=== "模型训练命令"
+
+    暂无
+
+=== "模型评估命令"
+
+    暂无
+
+=== "模型导出命令"
+
+    暂无
+
+=== "模型推理命令"
+
+    ``` sh
+    # Download sample input data
+    mkdir -p ./data
+    wget -nc https://paddle-org.bj.bcebos.com/paddlescience/models/Pangu/input_surface.npy -P ./data
+    wget -nc https://paddle-org.bj.bcebos.com/paddlescience/models/Pangu/input_upper.npy -P ./data
+
+    # Download pretrain model weight
+    mkdir -p ./inference
+
+    wget -nc https://paddle-org.bj.bcebos.com/paddlescience/models/Pangu/pangu_weather_1.onnx -P ./inference
+    wget -nc https://paddle-org.bj.bcebos.com/paddlescience/models/Pangu/pangu_weather_3.onnx -P ./inference
+    wget -nc https://paddle-org.bj.bcebos.com/paddlescience/models/Pangu/pangu_weather_6.onnx -P ./inference
+    wget -nc https://paddle-org.bj.bcebos.com/paddlescience/models/Pangu/pangu_weather_24.onnx -P ./inference
+
+    # 1h interval-time model inference
+    python predict.py INFER.export_path=inference/pangu_weather_1
+    # 3h interval-time model inference
+    python predict.py INFER.export_path=inference/pangu_weather_3
+    # 6h interval-time model inference
+    python predict.py INFER.export_path=inference/pangu_weather_6
+    # 24h interval-time model inference
+    python predict.py INFER.export_path=inference/pangu_weather_24
+    ```
+
+## 1. 背景简介
+
+盘古气象大模型(Pangu-Weather)是首个精度超过传统数值预报方法的 AI 方法，其提供了 1 小时间隔、3 小时间隔、6 小时间隔、24 小时间隔的预训练模型。1 小时 - 7 天预测精度均高于传统数值方法（即欧洲气象中心的 operational IFS），同时预测速度提升 10000 倍，可秒级完成对全球气象的预测，包括位势、湿度、风速、温度、海平面气压等。盘古气象大模型的水平空间分辨率达到 0.25°×0.25° ，时间分辨率为 1 小时，覆盖 13 层垂直高度，可以精准地预测细粒度气象特征。
+
+## 2. 模型原理
+
+本章节仅对盘古气象大模型的原理进行简单地介绍，详细的理论推导请阅读 [Pangu-Weather: A 3D High-Resolution System for Fast and Accurate Global Weather Forecast](https://arxiv.org/pdf/2211.02556)。
+
+模型的总体结构如图所示：
+
+待补充
+
+模型使用预训练权重推理，接下来将介绍模型的推理过程。
+
+## 3. 模型构建
+
+在该案例中，实现了 PanguWeatherPredictor用于ONNX模型的推理：
+
+``` py linenums="30" title="examples/pangu_weather/predict.py"
+--8<--
+examples/pangu_weather/predict.py:67:97
+--8<--
+```
+
+``` yaml linenums="15" title="examples/pangu_weather/conf/pangu_weather.yaml"
+--8<--
+examples/pangu_weather/conf/pangu_weather.yaml:29:44
+--8<--
+```
+
+其中，`input_file` 和 `input_surface_file` 分别代表网络模型输入的高空气象数据和地面气象。
+
+## 4. 结果可视化
+
+先将数据从 npy 转换为 NetCDF 格式，然后采用 ncvue 进行可视化
+
+1. 安装相关依赖
+```python
+pip install cdsapi netCDF4 ncvue
+```
+
+2. 使用脚本进行数据转换
+```python
+
+```
+
+3. 使用 ncvue 打开转换后的 NetCDF 文件
+
+## 5. 完整代码
+
+``` py linenums="1" title="examples/pangu_weather/predict.py"
+--8<--
+examples/pangu_weather/predict.py
+--8<--
+```
+
+## 6. 结果展示
+
+下图展示了模型的预测结果和真值结果。
+
+待补充
+
+## 7. 参考资料
+
+- [Pangu-Weather: A 3D High-Resolution System for Fast and Accurate Global Weather Forecast](https://arxiv.org/pdf/2211.02556)

examples/pangu_weather/convert_data.py

@@ -0,0 +1,142 @@
+# Copyright (c) 2025 PaddlePaddle Authors. All Rights Reserved.
+
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+
+#     http://www.apache.org/licenses/LICENSE-2.0
+
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+#
+# ref: https://github.com/HaxyMoly/Pangu-Weather-ReadyToGo/blob/main/forecast_decode_functions.py
+
+import os
+from os import path as osp
+from typing import Dict
+
+import hydra
+import netCDF4 as nc
+import numpy as np
+
+
+def convert_surface_data_to_nc(
+    surface_file: str, file_name: str, output_dir: str
+) -> None:
+    surface_data = np.load(surface_file)
+    mean_sea_level_pressure = surface_data[0]
+    u_component_of_wind_10m = surface_data[1]
+    v_component_of_wind_10m = surface_data[2]
+    temperature_2m = surface_data[3]
+
+    with nc.Dataset(
+        os.path.join(output_dir, file_name), "w", format="NETCDF4_CLASSIC"
+    ) as nc_file:
+        # Create dimensions
+        nc_file.createDimension("longitude", 1440)
+        nc_file.createDimension("latitude", 721)
+
+        # Create variables
+        nc_lon = nc_file.createVariable("longitude", np.float32, ("longitude",))
+        nc_lat = nc_file.createVariable("latitude", np.float32, ("latitude",))
+        nc_msl = nc_file.createVariable(
+            "mean_sea_level_pressure", np.float32, ("latitude", "longitude")
+        )
+        nc_u10 = nc_file.createVariable(
+            "u_component_of_wind_10m", np.float32, ("latitude", "longitude")
+        )
+        nc_v10 = nc_file.createVariable(
+            "v_component_of_wind_10m", np.float32, ("latitude", "longitude")
+        )
+        nc_t2m = nc_file.createVariable(
+            "temperature_2m", np.float32, ("latitude", "longitude")
+        )
+
+        # Set variable attributes
+        nc_lon.units = "degrees_east"
+        nc_lat.units = "degrees_north"
+        nc_msl.units = "Pa"
+        nc_u10.units = "m/s"
+        nc_v10.units = "m/s"
+        nc_t2m.units = "K"
+
+        # Write data to variables
+        nc_lon[:] = np.linspace(0.125, 359.875, 1440)
+        nc_lat[:] = np.linspace(90, -90, 721)
+        nc_msl[:] = mean_sea_level_pressure
+        nc_u10[:] = u_component_of_wind_10m
+        nc_v10[:] = v_component_of_wind_10m
+        nc_t2m[:] = temperature_2m
+
+
+def convert_upper_data_to_nc(upper_file: str, file_name: str, output_dir: str) -> None:
+    # Load the saved numpy arrays
+    upper_data = np.load(upper_file)
+    geopotential = upper_data[0]
+    specific_humidity = upper_data[1]
+    temperature = upper_data[2]
+    u_component_of_wind = upper_data[3]
+    v_component_of_wind = upper_data[4]
+
+    with nc.Dataset(
+        os.path.join(output_dir, file_name), "w", format="NETCDF4_CLASSIC"
+    ) as nc_file:
+        # Create dimensions
+        nc_file.createDimension("longitude", 1440)
+        nc_file.createDimension("latitude", 721)
+        nc_file.createDimension("level", 13)
+
+        # Create variables
+        nc_lon = nc_file.createVariable("longitude", np.float32, ("longitude",))
+        nc_lat = nc_file.createVariable("latitude", np.float32, ("latitude",))
+        nc_geopotential = nc_file.createVariable(
+            "geopotential", np.float32, ("level", "latitude", "longitude")
+        )
+        nc_specific_humidity = nc_file.createVariable(
+            "specific_humidity", np.float32, ("level", "latitude", "longitude")
+        )
+        nc_temperature = nc_file.createVariable(
+            "temperature", np.float32, ("level", "latitude", "longitude")
+        )
+        nc_u_component_of_wind = nc_file.createVariable(
+            "u_component_of_wind", np.float32, ("level", "latitude", "longitude")
+        )
+        nc_v_component_of_wind = nc_file.createVariable(
+            "v_component_of_wind", np.float32, ("level", "latitude", "longitude")
+        )
+
+        # Set variable attributes
+        nc_lon.units = "degrees_east"
+        nc_lat.units = "degrees_north"
+        nc_geopotential.units = "m"
+        nc_specific_humidity.units = "kg/kg"
+        nc_temperature.units = "K"
+        nc_u_component_of_wind.units = "m/s"
+        nc_v_component_of_wind.units = "m/s"
+        # Write data to variables
+        nc_lon[:] = np.linspace(0.125, 359.875, 1440)
+        nc_lat[:] = np.linspace(90, -90, 721)
+        nc_geopotential[:] = geopotential
+        nc_specific_humidity[:] = specific_humidity
+        nc_temperature[:] = temperature
+        nc_u_component_of_wind[:] = u_component_of_wind
+        nc_v_component_of_wind[:] = v_component_of_wind
+
+
+@hydra.main(version_base=None, config_path="./conf", config_name="pangu_weather.yaml")
+def main(cfg: Dict):
+    output_dir = cfg.output_dir
+
+    convert_surface_data_to_nc(
+        osp.join(output_dir, "output_surface.npy"), "output_surface.nc", output_dir
+    )
+    convert_upper_data_to_nc(
+        osp.join(output_dir, "output_upper.npy"), "output_upper.nc", output_dir
+    )
+
+
+if __name__ == "__main__":
+    main()

mkdocs.yml

@@ -104,6 +104,7 @@ nav:
         - VelocityGAN: zh/examples/velocity_gan.md
         - TGCN: zh/examples/tgcn.md
         - IOPS: zh/examples/iops.md
+        - Pang-Weather: zh/examples/pangu_weather.md
       - 化学科学(AI for Chemistry):
         - Moflow: zh/examples/moflow.md
         - IFM: zh/examples/ifm.md