Calling Python Script from SAS EM via SAS Code Node

[p0]

Hey everyone, 

I am interesting in running XGBOOST (and possibly CATBOOST and Light GBM) in SAS EM through a SAS Code node which will execute script in Python (where the actual models will be run).  I'm seeing that the SAS Code node may be used to call a Python script from the following video: https://video.sas.com/detail/videos/sas-enterprise-miner_/video/5483761790001/integrating-python-cod...

However, ultimately, I may be required to share the model with co-workers as a model package.  I'm not clear on how this would work if I am running a Python Script outside of SAS EM.  Would the other users just have to have python installed?  Would the model package somehow record the python script to run?  I'm pretty new at this technique so any and all help is appreciated!

Best, 

Nate

[AndrewZ]

My solution to put gradient boosting models into production in SAS is to use the gbm2sas function in the mlmeta R package. This function creates pure SAS code that runs without any dependencies on Python, and it does not even require SAS Enterprise Miner, SAS/IML, etc. 

It does not use LightGBM, Catboost, or XGBoost. It uses GBM. I recently did a small test here comparing GBM vs. LightBoost. LightGBM was much faster, but GBM made more accurate predictions. You can also speed up GBM in SAS using various tricks such as model tuning and parallel processing (yes, in SAS).