iterative
Connect metrics with model
One important thing about model registry is showing which metrics were changed and how after in the specific model version. Although you can produce metrics in DVC Pipeline, DVC doesn't provide a clear connection between models trained in Pipeline and metrics stored. E.g., you can have multiple models trained in one Pipeline, then you have no DVC-provided way to find out which metrics describe which model. Another case is when you produce a model without DVC pipeline. Would be good to specify metric values for a model. An important thing here is that this shouldn't be done after saving the model because model evaluation if often done when model is already saved. Thus, addition of some new data shouldn't change model's meta (because a model effectively doesn't change even if some new metrics are calculated). Right now we assume that model changes if mlem.yaml changes.
Pre-requisite for #98.
Currently MR is taking another direction and looks like this functionality is not needed in MLEM. Although something like collecting hardware reqs may be useful. Thus I'm closing this ticket.