I created a ML model with h2o's AutoML in Python and exported it as MOJO. I know I can run it again in Python when using the h2o package and an h2o server / the java application.
But I was wondering if and how I could convert/transform/use this model without using an h2o server. I only want to use h2o for the model generation, but not for running it.
Is this even possible, and if yes, how?
Thanks in advance, T0Bi
H2O models/MOJOs run on Java and need a JVM. Python will still need to run the job in Java.
If you don't want to initialize an H2O server, then the best way is to use Java directly and score with the MOJO. See more here.
Other option is using PySpark to score/predict without initializing an H2O server, but you need Spark set up in that case.