Building custom docker using spark and mleap

[prasadpande1990]

Hi,

I was referring to this repo to build a custom docker image that does the feature transformation and use lightGBM model for prediction.

I wanted to use mleap to serialize my feature transformation model which is basically a spark pipeline model.

I tried adding dependencies in my docker and tried training but while calling serializeToBundle method I’m getting following exception.

JavaPackage object is not callable

Is there a way to get around this issue.

All stackoverflow posts recommend adding mleap jars but I’m not sure how to add those jars to a sage maker custom docker.

Can you guide me with that?

Thank You,

Prasad