I am so excited to see that the long-planned release of the Makani library
github.com/NVIDIA/makani for
#AI based weather and climate modeling has finally happened!
It started with our FourCastNet work almost three years ago using the Fourier neural operator, where we showed for the first time that AI weather forecasting at high resolution (25km) was competitive with respect to traditional forecasts.
Since then, we demonstrated efficient scaling of our Fourier neural operator models that won the best paper award at PASC '23 and also developed the spherical Fourier neural operator that shows stable long rollouts in addition to medium-range forecasts. This is now running alongside numerical weather forecasts at
@ECMWF and the training code is now open source!