- one model per input row (probably the biggest use case)
- many models per input row
you must be consistent at training and prediction time, and you must be consistent with your use case. if you want both, create separate concordia instances (which is pretty easy)
you must be consistent at training and prediction time, and you must be consistent with your use case. if you want both, create separate concordia instances (which is pretty easy)