revert PR #3611 (#3786)

deepspeed/runtime/engine.py

@@ -1044,7 +1044,7 @@ class DeepSpeedEngine(Module):
             self.__check_params(self.module, torch.float)
 
         # zero.Init() handles device placement of model
-        if not (self.dont_change_device or is_zero3_model):
+        if not self.dont_change_device:
             self.module.to(self.device)
 
         # MoE related initialization
@@ -1080,7 +1080,7 @@ class DeepSpeedEngine(Module):
         self.expert_parallel_group = groups._get_expert_parallel_group_dict()
         self.expert_data_parallel_group = groups._get_expert_data_parallel_group_dict()
 
-        if not (self.amp_enabled() or is_zero3_model):
+        if not self.amp_enabled():
             self._broadcast_model()
 
     # check if parameters are duplicated in optimizer param_groups