The WeatherGenerator project is developing a machine learning-based Earth system model. It will be trained on a wide range of datasets, including reanalyses, forecast data and observations, to provide a robust and versatile model for the dynamics. Through this, it can be used for a wide-range of applications.
More details coming soon. Please open an issue if you are interested in using the model.
The model is currently being developed by the WeatherGenerator Consortium. If you want to engage, you are encouraged to contact us first by opening an issue on Github.
The main branch is the most stable version. If you are running experiments, you should use this branch.
The develop branch has the latest features. However, it is currently evolving at a fast pace. It should not be expected to have stable code or weight interfaces, or to be backward compatible.