Model Management/deployment

[abramvandergeest]

Current behavior: single model given for each ML inference activity.

Expected behavior: Tooling to allow for Soft/hard roll over between models. A way to have inference:

• do percentages of times used for one model or another
• allow multiple files be dropped and pick the newest

What is the motivation / use case for changing the behavior?

• Eventually ML models get stale and must be retrained
• A/B testing of models
• slow transition between models

Additional information you deem important (e.g. I need this tomorrow):