search

CliMA / OceanParameterizations.jl

Machine learning (and uncertainty quantification?) of climate model parameterizations using differentiable (and probabilistic?) programming.
MIT License
21 stars 2 forks source link

Consolidate free convection training pipeline and use proto-LESbrary.jl data #39

Closed ali-ramadhan closed 3 years ago

ali-ramadhan commented 3 years ago

This PR refactors the free convection training scripts into a single pipeline script appropriate for running on clusters for hyperparameter optimization/inference.

It uses DataDeps.jl to download the proto-LESbrary.jl data and GeoData.jl helps with data analysis. It also uses @xkykai's improved coarse graining for Face-centered fields. Animations are much better now (heat flux predictions and temperature evolution plotted in the same figure/animation).

This PR shouldn't touch the OceanParameterizations module. Maybe the coarse_grain functions would be useful, but I think we might want to develop some more general coarse graining functions first.

ali-ramadhan commented 3 years ago

This PR is dragging on so I might as well merge and continue working on a new branch.