Codes for "A Structured Graph Neural Network for Improving the Numerical Weather Prediction of Rainfall"
a) anaconda & jupyter notebook : https://www.anaconda.com/products/individual
b) PyTorch : https://pytorch.org/
c) deep graph libary : https://www.dgl.ai/
d) meteva : https://www.showdoc.com.cn/meteva/3975600102516402
Just run train_main.py , a single GIPMN model will begin training. There is a tiny dataset with 10 examples used for demonstrate the training process.
If you want to train a model for a specific rainfall level, for example, 10 mm, run train_main.py -- level_threshold 10 -- level_width 3, where the parameter "level_threshold" is the lower bound of the rainfall level, "level_width" is the rainfall range corresponding to probability of 0.1 ~ 0.9 .