Feedback on new ModelBuilder class for deploying PyMC models

I have been working on a new ModelBuilder class with @twiecki for PyMC which extends the current model class by adding a few more functionalities.
Here is a link for a blog post explaining the use of ModelBuilder class for deploying PyMC models
Please share what you think about this.


This looks really cool and useful.

I think that useful aspect about such a ModelBuilder class could be to enable users to integrate pymc models into scikit-learn pipeline to combine them with transformers and stuff.
To make that possible, the .fit() and .predict() would need to match the typical scikit-learn signature. Have you considered writing the function in that way?