search

ai2cm / fv3net

explore the FV3 data for parameterization
MIT License
16 stars 3 forks source link

Check into the consistency of the SklearnTransformer vs. Autoencoder with leading samples or single input. #2263

Closed frodre closed 1 year ago

frodre commented 1 year ago

I checked into this behavior. Currently the Autoencoder model fails when provided with 1D input. SkTransformer works given that it enforces 1D -> 2D in its implementation.

@AnnaKwa, perhaps we should use the ensure_sample_dim function for the Autoencoder as well?

notebook

AnnaKwa commented 1 year ago

Thanks for checking- yeah, that sounds good.