How can I fit lightGBM in spark2.4.5 and scala 2.1.1?

[WanBenLe]

Or how can I download mmlspark jar files and warppers for spark2.4.5 and scala 2.1.1?

AB#1989962

[github-actions[bot]]

Hey @WanBenLe :wave:! Thank you so much for reporting the issue/feature request :rotating_light:. Someone from SynapseML Team will be looking to triage this issue soon. We appreciate your patience.

[serena-ruan]

@WanBenLe Hey, spark2.4 is back in the days, and we haven't updated it for a long time. It's even before we rename to SynapseML, so my suggestion would be that maybe you could upgrade to spark3.1/spark3.2 which contains lots of new features :D If spark2.4 is the only choice, could you try install mmlspark version 1.0.0-rc3-67-659feb9c-SNAPSHOT and specify resolver as https://mmlspark.azureedge.net/maven? At the bottom of this website it elaborates how to install different SynapseML versions on different platforms as well.

[Usmaniatech]

Is mmlspark necessary to save/serialize the model? I am able to train the model but not able to save it.

[serena-ruan]

Is mmlspark necessary to save/serialize the model? I am able to train the model but not able to save it.

You should be able to save the model by calling 'model.write.save()' in Scala or 'model.write().save()' in Python.

[Usmaniatech]

Is mmlspark necessary to save/serialize the model? I am able to train the model but not able to save it.

You should be able to save the model by calling 'model.write.save()' in Scala or 'model.write().save()' in Python.

Does it support Java. I need to save it in java. I have no success so far.

[serena-ruan]

I think you can call Scala from Java and vice versa. If you've already got the trained model, you should be able to call write() method which generates the MLWriter, and then call .save() to save the model.