[REVIEW] RF: Variable binning and other minor refactoring

cpp/bench/sg/fil.cu

@@ -148,7 +148,7 @@ std::vector<Params> getInputs()
   p.rf = set_rf_params(10,                 /*max_depth */
                        (1 << 20),          /* max_leaves */
                        1.f,                /* max_features */
-                       32,                 /* n_bins */
+                       32,                 /* max_n_bins */
                        3,                  /* min_samples_leaf */
                        3,                  /* min_samples_split */
                        0.0f,               /* min_impurity_decrease */

cpp/bench/sg/rf_classifier.cu

@@ -95,7 +95,7 @@ std::vector<Params> getInputs()
   p.rf = set_rf_params(10,                  /*max_depth */
                        (1 << 20),           /* max_leaves */
                        0.3,                 /* max_features */
-                       32,                  /* n_bins */
+                       32,                  /* max_n_bins */
                        3,                   /* min_samples_leaf */
                        3,                   /* min_samples_split */
                        0.0f,                /* min_impurity_decrease */

cpp/bench/sg/rf_regressor.cu

@@ -95,7 +95,7 @@ std::vector<RegParams> getInputs()
   p.rf                          = set_rf_params(10,                 /*max_depth */
                        (1 << 20),          /* max_leaves */
                        0.3,                /* max_features */
-                       32,                 /* n_bins */
+                       32,                 /* max_n_bins */
                        3,                  /* min_samples_leaf */
                        3,                  /* min_samples_split */
                        0.0f,               /* min_impurity_decrease */

cpp/include/cuml/ensemble/randomforest.hpp

@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2019-2021, NVIDIA CORPORATION.
+ * Copyright (c) 2019-2022, NVIDIA CORPORATION.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -191,7 +191,7 @@ RF_metrics score(const raft::handle_t& user_handle,
 RF_params set_rf_params(int max_depth,
                         int max_leaves,
                         float max_features,
-                        int n_bins,
+                        int max_n_bins,
                         int min_samples_leaf,
                         int min_samples_split,
                         float min_impurity_decrease,

cpp/include/cuml/tree/decisiontree.hpp

@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2019-2021, NVIDIA CORPORATION.
+ * Copyright (c) 2019-2022, NVIDIA CORPORATION.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -40,9 +40,9 @@ struct DecisionTreeParams {
    */
   float max_features;
   /**
-   * Number of bins used by the split algorithm.
+   * maximum number of bins used by the split algorithm per feature.
    */
-  int n_bins;
+  int max_n_bins;
   /**
    * The minimum number of samples (rows) in each leaf node.
    */
@@ -74,7 +74,7 @@ struct DecisionTreeParams {
  * @param[in] cfg_max_depth: maximum tree depth; default -1
  * @param[in] cfg_max_leaves: maximum leaves; default -1
  * @param[in] cfg_max_features: maximum number of features; default 1.0f
- * @param[in] cfg_n_bins: number of bins; default 8
+ * @param[in] cfg_max_n_bins: maximum number of bins; default 128
  * @param[in] cfg_min_samples_leaf: min. rows in each leaf node; default 1
  * @param[in] cfg_min_samples_split: min. rows needed to split an internal node;
  *            default 2
@@ -84,13 +84,13 @@ struct DecisionTreeParams {
  *            i.e., GINI for classification or MSE for regression
  * @param[in] cfg_max_batch_size: Maximum number of nodes that can be processed
               in a batch. This is used only for batched-level algo. Default
-              value 128.
+              value 4096.
  */
 void set_tree_params(DecisionTreeParams& params,
                      int cfg_max_depth               = -1,
                      int cfg_max_leaves              = -1,
                      float cfg_max_features          = 1.0f,
-                     int cfg_n_bins                  = 128,
+                     int cfg_max_n_bins              = 128,
                      int cfg_min_samples_leaf        = 1,
                      int cfg_min_samples_split       = 2,
                      float cfg_min_impurity_decrease = 0.0f,

cpp/src/decisiontree/batched-levelalgo/bins.cuh

@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2019-2021, NVIDIA CORPORATION.
+ * Copyright (c) 2019-2022, NVIDIA CORPORATION.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -25,9 +25,9 @@ struct CountBin {
   HDI CountBin(int x_) : x(x_) {}
   HDI CountBin() : x(0) {}
 
-  DI static void IncrementHistogram(CountBin* hist, int nbins, int b, int label)
+  DI static void IncrementHistogram(CountBin* hist, int n_bins, int b, int label)
   {
-    auto offset = label * nbins + b;
+    auto offset = label * n_bins + b;
     CountBin::AtomicAdd(hist + offset, {1});
   }
   DI static void AtomicAdd(CountBin* address, CountBin val) { atomicAdd(&address->x, val.x); }
@@ -51,7 +51,7 @@ struct AggregateBin {
   HDI AggregateBin() : label_sum(0.0), count(0) {}
   HDI AggregateBin(double label_sum, int count) : label_sum(label_sum), count(count) {}
 
-  DI static void IncrementHistogram(AggregateBin* hist, int nbins, int b, double label)
+  DI static void IncrementHistogram(AggregateBin* hist, int n_bins, int b, double label)
   {
     AggregateBin::AtomicAdd(hist + b, {label, 1});
   }