Hello,
I have a question about best practice regarding registering a feature in Databricks feature store.
โ
Lets say that I create and register featuresโ during the EDA or experiment phase of a ML project.
Later the model is moving to production stage or other models are beginning to consume the features.
Is it best practice to then move the feature creation code into its own standalone Notebook?
โ
Related question:
A notebook registers features, but when the feature store generates those features for consumption , do all the cells in the associated notebook run , or can it determine only the code that relates specifically to the feature ETL?