Describe likely-used & "paper-used" parameters for executing steps

m2lines / gz21_ocean_momentum

Stochastic-Deep Learning Parameterization of Ocean Momentum Forcing

MIT License

5 stars 1 forks source link

Describe likely-used & "paper-used" parameters for executing steps #80

Closed raehik closed 9 months ago

raehik commented 1 year ago

The readme provides some detail on step parameters (data processing, training), plus provides a default command to run. But rationales, values you are likely to want to use, notes for unusual things (e.g. long_min=-280 because of how the CM2.6 dataset is recorded) aren't given in much detail.

It'd be great to discuss these parameters in a more high-level fashion in some documentation.

raehik commented 9 months ago

We point Arthur's 2021 paper in the readme, which covers these things in the appropriate detail, and general documentation is greatly improved in #97 , including instructions for reproducing paper results. I'm closing this as implemented.