Update Python doc. [skip ci] (#5517)

* Update doc for copying booster. [skip ci]

The issue is resolved in  #5312 .

* Add version for new APIs. [skip ci]

python-package/xgboost/core.py

@@ -1048,6 +1048,9 @@ class DeviceQuantileDMatrix(DMatrix):
     quantisation.
 
     You can construct DeviceQuantileDMatrix from cupy/cudf/dlpack.
+
+    .. versionadded:: 1.1.0
+
     """
 
     def __init__(self, data, label=None, weight=None, base_margin=None,
@@ -1197,7 +1200,10 @@ def __setstate__(self, state):
 
     def save_config(self):
         '''Output internal parameter configuration of Booster as a JSON
-        string.'''
+        string.
+
+        .. versionadded:: 1.0.0
+        '''
         json_string = ctypes.c_char_p()
         length = c_bst_ulong()
         _check_call(_LIB.XGBoosterSaveJsonConfig(
@@ -1208,7 +1214,10 @@ def save_config(self):
         return json_string
 
     def load_config(self, config):
-        '''Load configuration returned by `save_config`.'''
+        '''Load configuration returned by `save_config`.
+
+        .. versionadded:: 1.0.0
+        '''
         assert isinstance(config, str)
         _check_call(_LIB.XGBoosterLoadJsonConfig(
             self.handle,
@@ -1218,11 +1227,7 @@ def __copy__(self):
         return self.__deepcopy__(None)
 
     def __deepcopy__(self, _):
-        '''Return a copy of booster.  Caches for DMatrix are not copied so continue
-        training on copied booster will result in lower performance and
-        slightly different result.
-
-        '''
+        '''Return a copy of booster.'''
         return Booster(model_file=self)
 
     def copy(self):
@@ -1517,6 +1522,8 @@ def predict(self,
             Whether the prediction value is used for training.  This can effect
             `dart` booster, which performs dropouts during training iterations.
 
+            .. versionadded:: 1.0.0
+
         .. note:: Using ``predict()`` with DART booster
 
           If the booster object is DART type, ``predict()`` will not perform
@@ -1601,6 +1608,8 @@ def inplace_predict(self, data, iteration_range=(0, 0),
             booster.set_param({'predictor': 'cpu_predictor})
             booster.inplace_predict(numpy_array)
 
+        .. versionadded:: 1.1.0
+
         Parameters
         ----------
         data : numpy.ndarray/scipy.sparse.csr_matrix/cupy.ndarray/

python-package/xgboost/dask.py

@@ -126,6 +126,8 @@ class DaskDMatrix:
     the input data explicitly if you want to see actual computation of
     constructing `DaskDMatrix`.
 
+    .. versionadded:: 1.0.0
+
     Parameters
     ----------
     client: dask.distributed.Client
@@ -368,6 +370,8 @@ def _get_rabit_args(worker_map, client):
 def train(client, params, dtrain, *args, evals=(), **kwargs):
     '''Train XGBoost model.
 
+    .. versionadded:: 1.0.0
+
     Parameters
     ----------
     client: dask.distributed.Client
@@ -459,6 +463,8 @@ def predict(client, model, data, *args, missing=numpy.nan):
 
         Only default prediction mode is supported right now.
 
+    .. versionadded:: 1.0.0
+
     Parameters
     ----------
     client: dask.distributed.Client
@@ -582,6 +588,8 @@ def inplace_predict(client, model, data,
                     missing=numpy.nan):
     '''Inplace prediction.
 
+    .. versionadded:: 1.1.0
+
     Parameters
     ----------
     client: dask.distributed.Client