Add doc.

python-package/xgboost/core.py

@@ -1526,27 +1526,28 @@ def inplace_predict(self, data, iteration_range=(0, 0),
 
         Parameters
         ----------
-        data : numpy.ndarray/scipy.sparse.csr_matrix/cupy.ndarray
+        data : numpy.ndarray/scipy.sparse.csr_matrix/cupy.ndarray/
+               cudf.DataFrame/pd.DataFrame
             The input data, must not be a view for numpy array.  Set
             ``predictor`` to ``gpu_predictor`` for running prediction on CuPy
-            array.
+            array or CuDF DataFrame.
         iteration_range : tuple
             Specifies which layer of trees are used in prediction.  For
             example, if a random forest is trained with 100 rounds.  Specifying
-            `iteration_range=(10, 20)`, then only the forest built during [10,
+            `iteration_range=(10, 20)`, then only the forests built during [10,
             20) (open set) rounds are used in this prediction.
         predict_type : str
             * `value` Output model prediction values.
             * `margin` Output the raw untransformed margin value.
-        missing :
-            missing value of data.  Please refer to document for ``DMatrix``.
+        missing : float
+            Value in the input data which needs to be present as a missing
+            value.
 
         Returns
         -------
         prediction : numpy.ndarray/cupy.ndarray
-
-            The prediction result.  When input data is cupy array, prediction
-            result is also stored in a cupy array.
+            The prediction result.  When input data is on GPU, prediction
+            result is stored in a cupy array.
 
         '''
 

src/predictor/cpu_predictor.cc

@@ -48,6 +48,7 @@ struct SparsePageView {
 
   explicit SparsePageView(SparsePage const *p)
       : page{p}, base_rowid{page->base_rowid} {
+    // Pull to host before entering omp block, as this is not thread safe.
     page->data.HostVector();
     page->offset.HostVector();
   }
@@ -110,7 +111,6 @@ void PredictBatchKernel(DataView batch, std::vector<bst_float> *out_preds,
   constexpr int kUnroll = 8;
   const auto nsize = static_cast<bst_omp_uint>(batch.Size());
   const bst_omp_uint rest = nsize % kUnroll;
-  // Pull to host before entering omp block, as this is not thread safe.
   if (nsize >= kUnroll) {
 #pragma omp parallel for schedule(static)
     for (bst_omp_uint i = 0; i < nsize - rest; i += kUnroll) {