Geek Answers Handbook

How to save / export Spark ML Lib model in PMML?

I would like to train the model using Spark ML Lib, but then I can export the model in a platform-agnostic format. Essentially, I want to separate how models are created and consumed.

My reason is that you want to untie, so that I can deploy the model to other projects. For instance:.

  • Use the model to make forecasts in a separate, stand-alone program that is independent of Spark for evaluation.
  • Use the model with existing projects, such as OpenScoring, and provide APIs that can use the model.
  • Upload your existing model back to Spark to predict high throughput.

Has anyone done something similar with the Spark ML Lib?

+5
source share
1 answer

Spark 1.4 version now has support for this. See the latest documentation . Not all models are available (see Support (see JIRA SPARK-4587 issue ).

HTHS

+4
source

More articles:


All Articles