Is your feature request related to a problem? Please describe.
For inference and deployment of models in production, I would like to create serializable models. Such models allow us to run inference without the code to build the architecture; furthermore, input preprocessing functions can be scripted with the models hence minimizing input drift when using models.
Describe the solution you'd like
A base class for scripting models during training; the function call is optional.
Is your feature request related to a problem? Please describe. For inference and deployment of models in production, I would like to create serializable models. Such models allow us to run inference without the code to build the architecture; furthermore, input preprocessing functions can be scripted with the models hence minimizing input drift when using models.
Describe the solution you'd like A base class for scripting models during training; the function call is optional.
Describe alternatives you've considered None
Additional context None