Is the exported model able to run in python environment instead of Java/Scala environment?

combust / mleap

MLeap: Deploy ML Pipelines to Production

https://combust.github.io/mleap-docs/

Apache License 2.0

1.5k stars 312 forks source link

Is the exported model able to run in python environment instead of Java/Scala environment? #491

Open Exlsunshine opened 5 years ago

Exlsunshine commented 5 years ago

I am trying to use Mleap to avoid Spark during prediction. Currently, I can export the Spark model to Mleap format, but I cannot find any documents that can tell me how to use the exported model without spark dependency, I do find some wikis that describe how to do it in Java/Scala, but I wonder is the exported model able to run in python environment instead of Java/Scala environment?

ancasarb commented 5 years ago

Currently, mleap-runtime can only be invoked from Scala/Java/JVM based language. What you can do though, is to have either a grpc client or http client in Python, calling a scoring service running in Java/Scala. You can start by taking a look at the docker images with the spring boot/grpc services we released recently.

yairdata commented 5 years ago

@Exlsunshine from your issue i understand that you were able to use mleap with pyspark , can you share the versions & installation details , there are few issues open for the integration and it doesn't seem to work by the documentation (mleap 0.13 & spark 2.4.0./2.3.0 for example)