To support transfer learning allow models to depend on other models
inherit the original pipeline since the feature inputs will be the same
config should be the incremental config and freezing/fine tuning info to reproduce
*can also use this to support online learning (each model is an incremental version of a previous one with configs indicating the changes and data mapping to the new dataset input)
To support transfer learning allow models to depend on other models
*can also use this to support online learning (each model is an incremental version of a previous one with configs indicating the changes and data mapping to the new dataset input)