!8993 fix bug

From: @jinyaohui
Reviewed-by: @kingxian,@zh_qh
Signed-off-by: @kingxian

mindspore/common/api.py

@@ -379,16 +379,15 @@ class _Executor:
         auto_split_param_names = (param_name for param_name in auto_split_params)
         return auto_split_param_names
 
-    def _build_broadcast_graph(self, obj, broadcast_params, broadcast_phase):
+    def _build_broadcast_graph(self, broadcast_params_dict, broadcast_phase):
         """Build broadcast graph."""
         from mindspore.nn.wrap.cell_wrapper import _BroadCastCell
 
-        _broadcast_net = _BroadCastCell(broadcast_params)
+        _broadcast_net = _BroadCastCell(broadcast_params_dict.values())
         _broadcast_net.phase = broadcast_phase
         broadcasted_params = _broadcast_net()
-        parameters_broadcast_dict = obj.parameters_broadcast_dict()
-        for param_name, param in zip(parameters_broadcast_dict, broadcasted_params):
-            parameters_broadcast_dict[param_name].set_data(param)
+        for param_name, param in zip(broadcast_params_dict.keys(), broadcasted_params):
+            broadcast_params_dict[param_name].set_data(param)
 
     def _set_dataset_mode(self, args_list):
         """set dataset mode."""
@@ -476,10 +475,15 @@ class _Executor:
             auto_split_param_names = []
             if auto_parallel_mode:
                 auto_split_param_names = self._get_auto_split_param_names(obj.parameter_layout_dict)
-            broadcast_params = [param for param_name, param in obj.parameters_broadcast_dict().items() if
-                                param_name not in auto_split_param_names]
-            broadcast_phase = "broadcast_subgraph" + "." + str(obj.create_time)
-            self._build_broadcast_graph(obj, broadcast_params, broadcast_phase)
+
+            broadcast_params_dict = obj.parameters_broadcast_dict()
+            if auto_split_param_names and broadcast_params_dict:
+                broadcast_params_dict = OrderedDict()
+                for param_name, param in obj.parameters_broadcast_dict().items():
+                    if param_name not in auto_split_param_names:
+                        broadcast_params_dict[param_name] = param
+            broadcast_phase = "_broadcast_subgraph" + "." + str(obj.create_time)
+            self._build_broadcast_graph(broadcast_params_dict, broadcast_phase)
             self.compile_cache[phase] = broadcast_phase
 
         return phase, True

mindspore/train/model.py

@@ -564,7 +564,6 @@ class Model:
             raise ValueError("The sink_size must be -1 or positive, but got sink_size {}.".format(sink_size))
 
         _device_number_check(self._parallel_mode, self._device_number)
-        _parameter_broadcast_check(self._parallel_mode, self._parameter_broadcast)
 
         self._train(epoch,
                     train_dataset,