Is your feature request related to a problem? Please describe.
Currently when a user runs get_historical_features() Feast will upload or create an entity dataframe in the offline store. Feast will then build a training dataset and return that to the user. Feast does not clean up the temporary tables, but often sets an expiry. This means that a data warehouse constantly has a set of temporary tables waiting to expire.
Describe the solution you'd like
Delete entity dataframe tables after a training dataset has been created (as well as other temporary tables in the creation process).
Is your feature request related to a problem? Please describe.
Currently when a user runs
get_historical_features()Feast will upload or create an entity dataframe in the offline store. Feast will then build a training dataset and return that to the user. Feast does not clean up the temporary tables, but often sets an expiry. This means that a data warehouse constantly has a set of temporary tables waiting to expire.Describe the solution you'd like
Delete entity dataframe tables after a training dataset has been created (as well as other temporary tables in the creation process).