make __fake_vars and _double_grads to lazy

python/paddle/fluid/dygraph/dygraph_to_static/partial_program.py

@@ -148,15 +148,7 @@ def __init__(self, main_program, inputs, outputs, parameters=None,
 
         self._origin_main_program = self._verify_program(main_program)
         self._tmp_scope_vec = self._create_scope_vec()
-        # 量化单测里调用jit.save时，save接口会整体切换到静态图模式，在save下构建partial_program时，_create_fake_var调用core.eager.Tensor会段错误
-        # 1. 询问jiabin看构造eager.Tensor是否可以在静态图下进行
-        # 2. _create_fake_var内new Tnesor时主动切换回动态图
-        # 3. save接口内，构建partial_progarm时切回动态图
-        # 4. 将__init__中new eager.Tensor的相关调用移动到__call__中
-        # A fake_var to handle empty input or output
-        self.__fake_vars = _create_fake_var()
         # Set default mode to train
-        self._double_grads = self._get_double_grads(self._origin_main_program)
         self.training = True
 
         custom_white_list, custom_black_list = None, None
@@ -168,6 +160,14 @@ def __init__(self, main_program, inputs, outputs, parameters=None,
             custom_white_list=custom_white_list,
             custom_black_list=custom_black_list)
 
+    @LazyInitialized
+    def __fake_vars(self):
+        return _create_fake_var()
+
+    @LazyInitialized
+    def _double_grads(self):
+        return self._get_double_grads(self._origin_main_program)
+
     @LazyInitialized
     def _infer_program(self):
         """