cobra
cobra copied to clipboard
Append model pipeline dict with relevant preprocessor pipelines for quicker scoring
I would like to append the model pipeline dict with relevant preprocessor pipelines. Some kind of "export model for scoring" possibility. The model.serialize JSON would then only contain the prepocessor pipelines of the variables actually used. Or maybe it shoud be an new JSON, I don't know.
I know i can use use the preprocessor JSON to transform all the variables, but for scoring on a new dataset, that's overkill. It would be very convinient to just having to run cobra.score_model(new_data,total_pipeline) :-)
PS: there is a discrepancy in naming between preprocessor.serialize_pipeline() and model.serialize()