Fix spelling in documents (#6948)

* Update roxygen2 doc.

Co-authored-by: fis <jm.yuan@outlook.com>

src/tree/constraints.h

@@ -17,7 +17,7 @@ namespace xgboost {
 /*!
  * \brief Feature interaction constraint implementation for CPU tree updaters.
  *
- * The interface is similiar to the one for GPU Hist.
+ * The interface is similar to the one for GPU Hist.
  */
 class FeatureInteractionConstraintHost {
  protected:

src/tree/gpu_hist/evaluate_splits.cu

@@ -125,7 +125,7 @@ struct UpdateNumeric {
                              EvaluateSplitInputs<GradientSumT> const &inputs,
                              DeviceSplitCandidate *best_split) {
     // Use pointer from cut to indicate begin and end of bins for each feature.
-    uint32_t gidx_begin = inputs.feature_segments[fidx];  // begining bin
+    uint32_t gidx_begin = inputs.feature_segments[fidx];  // beginning bin
     int split_gidx = (scan_begin + threadIdx.x) - 1;
     float fvalue;
     if (split_gidx < static_cast<int>(gidx_begin)) {
@@ -152,7 +152,7 @@ __device__ void EvaluateFeature(
     TempStorageT* temp_storage         // temp memory for cub operations
 ) {
   // Use pointer from cut to indicate begin and end of bins for each feature.
-  uint32_t gidx_begin = inputs.feature_segments[fidx];  // begining bin
+  uint32_t gidx_begin = inputs.feature_segments[fidx];  // beginning bin
   uint32_t gidx_end =
       inputs.feature_segments[fidx + 1];  // end bin for i^th feature
   auto feature_hist = inputs.gradient_histogram.subspan(gidx_begin, gidx_end - gidx_begin);

src/tree/gpu_hist/gradient_based_sampler.cuh

@@ -124,7 +124,7 @@ class ExternalMemoryGradientBasedSampling : public SamplingStrategy {
  * Processing Systems (pp. 3146-3154).
  * \see Zhu, R. (2016). Gradient-based sampling: An adaptive importance sampling for least-squares.
  * In Advances in Neural Information Processing Systems (pp. 406-414).
- * \see Ohlsson, E. (1998). Sequential poisson sampling. Journal of official Statistics, 14(2), 149.
+ * \see Ohlsson, E. (1998). Sequential Poisson sampling. Journal of official Statistics, 14(2), 149.
  */
 class GradientBasedSampler {
  public:

src/tree/gpu_hist/histogram.cu

@@ -17,7 +17,7 @@
 
 namespace xgboost {
 namespace tree {
-// Following 2 functions are slightly modifed version of fbcuda.
+// Following 2 functions are slightly modified version of fbcuda.
 
 /* \brief Constructs a rounding factor used to truncate elements in a sum such that the
    sum of the truncated elements is the same no matter what the order of the sum is.

src/tree/param.h

@@ -76,7 +76,7 @@ struct TrainParam : public XGBoostParameter<TrainParam> {
   // the criteria to use for ranking splits
   std::string split_evaluator;
 
-  // ------ From cpu quantile histogram -------.
+  // ------ From CPU quantile histogram -------.
   // percentage threshold for treating a feature as sparse
   // e.g. 0.2 indicates a feature with fewer than 20% nonzeros is considered sparse
   double sparse_threshold;
@@ -316,7 +316,7 @@ XGBOOST_DEVICE inline T CalcGain(const TrainingParams &p, StatT stat) {
   return CalcGain(p, stat.GetGrad(), stat.GetHess());
 }
 
-// Used in gpu code where GradientPair is used for gradient sum, not GradStats.
+// Used in GPU code where GradientPair is used for gradient sum, not GradStats.
 template <typename TrainingParams, typename GpairT>
 XGBOOST_DEVICE inline float CalcWeight(const TrainingParams &p, GpairT sum_grad) {
   return CalcWeight(p, sum_grad.GetGrad(), sum_grad.GetHess());
@@ -484,7 +484,7 @@ using SplitEntry = SplitEntryContainer<GradStats>;
 
 /*
  * \brief Parse the interaction constraints from string.
- * \param constraint_str String storing the interfaction constraints:
+ * \param constraint_str String storing the interaction constraints:
  *
  *  Example input string:
  *

src/tree/tree_model.cc

@@ -157,7 +157,7 @@ TreeGenerator* TreeGenerator::Create(std::string const& attrs, FeatureMap const&
   if (pos != std::string::npos) {
     name = attrs.substr(0, pos);
     params = attrs.substr(pos+1, attrs.length() - pos - 1);
-    // Eliminate all occurances of single quote string.
+    // Eliminate all occurrences of single quote string.
     size_t pos = std::string::npos;
     while ((pos = params.find('\'')) != std::string::npos) {
       params.replace(pos, 1, "\"");
@@ -1069,7 +1069,7 @@ void RegTree::CalculateContributionsApprox(const RegTree::FVec &feat,
 // Used by TreeShap
 // data we keep about our decision path
 // note that pweight is included for convenience and is not tied with the other attributes
-// the pweight of the i'th path element is the permuation weight of paths with i-1 ones in them
+// the pweight of the i'th path element is the permutation weight of paths with i-1 ones in them
 struct PathElement {
   int feature_index;
   bst_float zero_fraction;
@@ -1123,7 +1123,7 @@ void UnwindPath(PathElement *unique_path, unsigned unique_depth,
   }
 }
 
-// determine what the total permuation weight would be if
+// determine what the total permutation weight would be if
 // we unwound a previous extension in the decision path
 bst_float UnwoundPathSum(const PathElement *unique_path, unsigned unique_depth,
                          unsigned path_index) {

src/tree/updater_basemaker-inl.h

@@ -196,8 +196,8 @@ class BaseMaker: public TreeUpdater {
     }
   }
   /*!
-   * \brief this is helper function uses column based data structure,
-   *        reset the positions to the lastest one
+   * \brief This is a helper function that uses a column based data structure
+   *        and reset the positions to the latest one
    * \param nodes the set of nodes that contains the split to be used
    * \param p_fmat feature matrix needed for tree construction
    * \param tree the regression tree structure

src/tree/updater_gpu_hist.cu

@@ -549,7 +549,7 @@ struct GPUHistMakerDevice {
       bst_float weight = evaluator.CalcWeight(
           pos, param_d, GradStats{d_node_sum_gradients[pos]});
       static_assert(!std::is_const<decltype(out_preds_d)>::value, "");
-      auto v_predt = out_preds_d;  // for some reaon out_preds_d is const by both nvcc and clang.
+      auto v_predt = out_preds_d;  // for some reason out_preds_d is const by both nvcc and clang.
       v_predt[d_ridx[local_idx]] += weight * param_d.learning_rate;
     });
     row_partitioner.reset();

src/tree/updater_histmaker.cc

@@ -401,7 +401,7 @@ class CQHistMaker: public HistMaker {
     for (auto& sketch : sketchs_) {
       sketch.Init(info.num_row_, this->param_.sketch_eps);
     }
-    // intitialize the summary array
+    // initialize the summary array
     summary_array_.resize(sketchs_.size());
     // setup maximum size
     unsigned max_size = this->param_.MaxSketchSize();
@@ -409,7 +409,7 @@ class CQHistMaker: public HistMaker {
       summary_array_[i].Reserve(max_size);
     }
     {
-      // get smmary
+      // get summary
       thread_sketch_.resize(omp_get_max_threads());
 
       // TWOPASS: use the real set + split set in the column iteration.

src/tree/updater_quantile_hist.h

@@ -441,7 +441,7 @@ class QuantileHistMaker: public TreeUpdater {
     std::unique_ptr<ExpandQueue> qexpand_loss_guided_;
     std::vector<ExpandEntry> qexpand_depth_wise_;
     // key is the node id which should be calculated by Subtraction Trick, value is the node which
-    // provides the evidence for substracts
+    // provides the evidence for subtraction
     std::vector<ExpandEntry> nodes_for_subtraction_trick_;
     // list of nodes whose histograms would be built explicitly.
     std::vector<ExpandEntry> nodes_for_explicit_hist_build_;

src/tree/updater_refresh.cc

@@ -123,7 +123,7 @@ class TreeRefresher: public TreeUpdater {
     // start from groups that belongs to current data
     auto pid = 0;
     gstats[pid].Add(gpair[ridx]);
-    // tranverse tree
+    // traverse tree
     while (!tree[pid].IsLeaf()) {
       unsigned split_index = tree[pid].SplitIndex();
       pid = tree.GetNext(pid, feat.GetFvalue(split_index), feat.IsMissing(split_index));