Create training app and models

This is the first issue related to ML training in Arkindex !

We need to create a new django application named training, and then add 2 models:

• training.Model will define an evolving Machine learning model

  • UUID PK
  • CharField name, unique
  • description textfield
  • public boolean (default False)
  • created/update date
  • M2M workers towards dataimport.Worker named compatible_workers

• training.ModelVersion will define a specific model version, that's where we control the model data

  • UUID PK
  • ForeignKey toward training.Model named model, related name is versions
  • nullable ForeignKey toward training.ModelVersion named parent, related name is children
  • description textfield
  • nullable charfield tag (unique together with FK model)
  • charfield hash (of the archive)
  • positive integer size (of the archive)
  • enum state
    • created
    • available (and thus checked by the backend)
    • error
  • JsonField configuration to store dict of parameters, provided by the ML dev
  • created/update date

graph TD

    Model --> ModelVersion
    Model --> Right
    Model --> ModelWorker --> Worker
    ModelVersion -- parent --> ModelVersion

Please make sure both models are available through the admin interface:

• Model with lists: name, created date (search on name & description)
• ModelVersion with lists: PK, model name, tag, size, state (filter on model name and state)