ADR: Object Saving and Loading with DataFlows

[johnandersen777]

This is a follow on to 0004-Model-Saving-and-Loading. This document outlines our
plan to implement a generic saving and load mechanism for all
:py:class:BaseConfigurable <dffml.base.BaseConfigurable> objects.
See the 2021-08-10 Weekly Sync Meeting recording for more details.

save and load of any object

• save

  • Takes the object and (a dataflow or a location or an operation)

    • Point here is to allow user control over how their object gets saved.

      • If a location is given we use the existing model archive dataflow creation

      • If operation is given, execute single operation dataflow

        • One could execute run_dataflow if one wanted configure different orchestrator or networks

    • Pass as inputs to the dataflow

      • plugin type (model)

      • which plugin (scikitlr)

      • the config (pass through :py:func:export <dffml.util.data.export>
        first).