Prevent training without setting up caches.

doc/python/python_intro.rst

@@ -161,6 +161,10 @@ A saved model can be loaded as follows:
   bst = xgb.Booster({'nthread': 4})  # init model
   bst.load_model('model.bin')  # load data
 
+Methods including `update` and `boost` from `xgboost.Booster` are designed for
+internal usage only.  The wrapper function `xgboost.train` does some
+pre-configuration including setting up caches and some other parameters.
+
 Early Stopping
 --------------
 If you have a validation set, you can use early stopping to find the optimal number of boosting rounds.
@@ -215,4 +219,3 @@ When you use ``IPython``, you can use the :py:meth:`xgboost.to_graphviz` functio
 .. code-block:: python
 
   xgb.to_graphviz(bst, num_trees=2)
-

python-package/xgboost/core.py

@@ -1041,8 +1041,8 @@ def set_param(self, params, value=None):
             _check_call(_LIB.XGBoosterSetParam(self.handle, c_str(key), c_str(str(val))))
 
     def update(self, dtrain, iteration, fobj=None):
-        """
-        Update for one iteration, with objective function calculated internally.
+        """Update for one iteration, with objective function calculated
+        internally.  This function should not be called directly by users.
 
         Parameters
         ----------
@@ -1052,6 +1052,7 @@ def update(self, dtrain, iteration, fobj=None):
             Current iteration number.
         fobj : function
             Customized objective function.
+
         """
         if not isinstance(dtrain, DMatrix):
             raise TypeError('invalid training matrix: {}'.format(type(dtrain).__name__))
@@ -1066,8 +1067,9 @@ def update(self, dtrain, iteration, fobj=None):
             self.boost(dtrain, grad, hess)
 
     def boost(self, dtrain, grad, hess):
-        """
-        Boost the booster for one iteration, with customized gradient statistics.
+        """Boost the booster for one iteration, with customized gradient
+        statistics.  Like :func:`xgboost.core.Booster.update`, this
+        function should not be called directly by users.
 
         Parameters
         ----------
@@ -1077,6 +1079,7 @@ def boost(self, dtrain, grad, hess):
             The first order of gradient.
         hess : list
             The second order of gradient.
+
         """
         if len(grad) != len(hess):
             raise ValueError('grad / hess length mismatch: {} / {}'.format(len(grad), len(hess)))

src/learner.cc

@@ -703,6 +703,8 @@ class LearnerImpl : public Learner {
     if (num_feature > mparam_.num_feature) {
       mparam_.num_feature = num_feature;
     }
+    CHECK_NE(mparam_.num_feature, 0)
+        << "0 feature is supplied.  Are you using raw Booster?";
     // setup
     cfg_["num_feature"] = common::ToString(mparam_.num_feature);
     CHECK(obj_ == nullptr && gbm_ == nullptr);