Improve the docstring in Experiment

PiperOrigin-RevId: 156097464

Author: Jianwei Xie

tensorflow/contrib/learn/python/learn/experiment.py

@@ -511,6 +511,8 @@ def continuous_train_and_eval(self,
     (via constructor). The model will be first trained for
     `train_steps_per_iteration`, and then be evaluated in turns.
 
+    This method is intended for single machine usage.
+
     This differs from `train_and_evaluate` as follows:
       1. The procedure will have train and evaluation in turns. The model
       will be trained for a number of steps (usuallly smaller than `train_steps`