Add doc (#13765)

test=develop

Author: chengduo

paddle/fluid/pybind/pybind.cc

@@ -620,7 +620,23 @@ All parameter, weight, gradient are variables in Paddle.
 
   // -- python binds for parallel executor.
   py::class_<ParallelExecutor> pe(m, "ParallelExecutor");
-  py::class_<ExecutionStrategy> exec_strategy(pe, "ExecutionStrategy");
+  py::class_<ExecutionStrategy> exec_strategy(pe, "ExecutionStrategy", R"DOC(
+    ExecutionStrategy allows the user to more preciously control how to run
+    the program in ParallelExecutor by setting the property.
+
+    The available properties include:
+        use_cuda (bool): Whether to use CUDA or not. Default True.
+        num_threads (int): The number of threads that used to run the
+            operators in ParallelExecutor. If it is not set, it will be
+            set in ParallelExecutor according to the device count.
+            Default 0.
+        allow_op_delay (bool): Whether to delay the communication operators
+            to run. Default False.
+        num_iteration_per_drop_scope (int): how many iterations between
+            the two dropping local scopes. Default 100.
+
+        )DOC");
+
   exec_strategy.def(py::init())
       .def_property(
           "num_threads",
@@ -658,7 +674,25 @@ All parameter, weight, gradient are variables in Paddle.
                                   : ExecutionStrategy::kDefault;
       });
 
-  py::class_<BuildStrategy> build_strategy(pe, "BuildStrategy");
+  py::class_<BuildStrategy> build_strategy(pe, "BuildStrategy", R"DOC(
+    BuildStrategy allows the user to more preciously control how to
+    build the SSA Graph in ParallelExecutor by setting the property.
+
+    The available properties include:
+        reduce_strategy (str): There are two reduce strategies, 'AllReduce'
+            and 'Reduce'. If you want that all parameters will be optimized
+            on all devices, you can choose 'AllReduce'; if you choose
+            'Reduce', all parameters will be evenly allocated to different
+            devices for optimization, and then broadcast the optimized
+            parameter to other devices. Default 'AllReduce'.
+        gradient_scale_strategy (str): There are two ways of defining loss@grad,
+            'CoeffNumDevice' and 'Customized'. By default, ParallelExecutor
+            sets the loss@grad according to the number of devices. If you want
+            to customize loss@grad, you can choose 'Customized'.
+            Default 'CoeffNumDevice'.
+        debug_graphviz_path (str): Whether to write the SSA Graph to file in the
+            form of graphviz. It is useful for debugging. Default "".
+)DOC");
 
   py::enum_<BuildStrategy::ReduceStrategy>(build_strategy, "ReduceStrategy")
       .value("Reduce", BuildStrategy::ReduceStrategy::kReduce)