Closed Mantj closed 4 years ago
The pipeline wrapper is required, because it holds all sorts of supporting metadata about the learning process (eg. column names).
It should be possible to bypass this "convenience layer", and work directly with lower JPMML-SparkML library layers and APIs. However, I don't have the time to guide you through there.
I have a GBTClassificationModel, and I want to save it as pmml, but not put it into pipeline, what can I do? thanks~