Evaluating on a subset of trees

#1

Is there currently a way to evaluate on a subset of the trained trees?

#2

You can set ntree_limit to evaluate first ntree_limit trees. It is currently not possible to evaluate an arbitrary subset of trees.

#3

Cool, thanks! Is there a way to evaluate on the first n trees in the Scala Spark version? I can’t seem to find a way to do so…

#4

@gracelam The treeLimit parameter is in the works for XGBoost4J-Spark: https://github.com/dmlc/xgboost/pull/3532