Use encoder/decoder call directly

ai2cm / fv3net

explore the FV3 data for parameterization

MIT License

16 stars 3 forks source link

Use encoder/decoder call directly #2366

Closed AnnaKwa closed 1 year ago

AnnaKwa commented 1 year ago

Making a separate PR using @frodre 's commit that greatly speeds up training. Directly calling the encoder vs .predict reduces batch fitting time by >2-10x.