Questions about RAPIDS

Accelerated Computing GPU-Accelerated Libraries
#1

  1. After XGBoost being trained, how to export trained model for future use? Can such model be converted into model used by Scikit-learn?

  2. How to do influence by using model trained by cuML?

  3. in RAPIDS presentation slide, there is no more example/samples depicting how to evaluate and visualize. Here visualization means using python’s libraies like matplotlib?

  4. When deploying, can exported model be used on traditional CPU server(on which scikit-learn is installed)?

#2
  1. I use GTX 1070, on which there is only 8G device memory. DASK worker crashes even I set end_year=2001 and part_count=1. How to configure memory usage limitation by using RMM? Any way to resolve this?