Initial commit

Author: avolkov-intel

sklbench/benchmarks/sklearn_estimator.py

@@ -134,6 +134,16 @@ def get_subset_metrics_of_estimator(
                 and isinstance(iterations[0], Union[Numeric, NumpyNumeric].__args__)
             ):
                 metrics.update({"iterations": int(iterations[0])})
+        if hasattr(estimator_instance, "estimators_"):
+            estimators_with_trees = [
+                t
+                for t in estimator_instance.estimators_
+                if hasattr(t, "tree_") and hasattr(t.tree_, "node_count")
+            ]
+            if estimators_with_trees:
+                metrics["n_nodes"] = sum(
+                    t.tree_.node_count for t in estimators_with_trees
+                )
     if task == "classification":
         y_pred = convert_to_numpy(estimator_instance.predict(x))
         metrics.update(

sklbench/report/implementation.py

@@ -71,6 +71,8 @@
         # NB: 'n_clusters' is parameter of KMeans while
         # 'clusters' is number of computer clusters by DBSCAN
         "clusters",
+        # tree ensembles
+        "n_nodes",
     ],
     "incomparable": [
         "1st-mean run ratio",