AI/ML Model Definitions and Training

No-Code Platform
1
  1. Do we create the Model Definitions in dev and then train the models in test?
  2. If we ship the models to a customer, do we need to train them in dev before shipping or are they trained again in the customer environment once shipped? We need to understand the concept.
2

Hi Jaco,

The trained ML model “instance” is tenant data so you will need to go through the model definition training process in a given tenant before you can run inference against it that tenant.

So for your first question, yes you will need to train the model in test if you want to run inference against it there.

When shipping a model, there is no automatic process to train when models move between tenants or environments, so it will need to be trained in the customer environment/tenant in order to run inference against it.

1 Like
3

Thanks Jeremy that makes sense. Will reach out if there are any other questions as we go through the process.