NSF NCAR Community Research Earth Digital Intelligence Twin (CREDIT) FuXi 6-Hour Model Weights and Configuration
This repository contains the PyTorch checkpoint weights and data/model configuration files for the CREDIT WXFormer 6-hour model. More information about the training and verification of this model can be found in the Schreck et al. (2024) ArXiv preprint.
Data Access
Our model is trained on ERA5 Reanalysis Data from a subset of 16 of the 137 hybrid sigma-pressure vertical levels. The raw data are available on the NSF NCAR Research Data Archive. Processed data can be accessed on the CREDIT ERA5 Zarr Files globus collection.
Running the Model
To run the model, first install the CREDIT package from github.
Modify the paths in the finetune_final/model_predict.yml configuration file to point to the appropriate ERA5 and scaler directories.
Inference Providers
NEW
This model is not currently available via any of the supported third-party Inference Providers, and
HF Inference API was unable to determine this model's library.