Fix a bug for Parameter

1. Parameter's init_data() should have no effect if default_input already set;
2. This bug is introduced by 'decouple ParamValue from python';
3. An unit test case is added to ensure the right init_data() behavior.

mindspore/common/parameter.py

@@ -210,7 +210,6 @@ class Parameter:
 
     def set_parameter_data(self, data):
         """Set `default_input` of current `Parameter`."""
-        self.init_mode = None
         if isinstance(data, bool):
             raise ValueError('Parameter data can not be `bool`')
         if isinstance(data, Tensor):
@@ -243,7 +242,8 @@ class Parameter:
             set_sliced (bool): True if should set parameter sliced after init the data of initializer.
                 Default: False.
         """
-        if self.init_mode is None:
+        if isinstance(self.default_input, Tensor):
+            # skip if data already initialized.
             return
         if layout is not None:
             if not isinstance(layout, list):

tests/ut/python/nn/test_parameter.py

@@ -134,3 +134,19 @@ def test_check_str_by_regular():
         _check_str_by_regular(str5)
     with pytest.raises(ValueError):
         _check_str_by_regular(str6)
+
+def test_parameter_lazy_init():
+    # Call init_data() without set default_input.
+    para = Parameter(initializer('ones', [1, 2, 3], mstype.float32), 'test1')
+    assert not isinstance(para.default_input, Tensor)
+    para.init_data()
+    assert isinstance(para.default_input, Tensor)
+    assert np.array_equal(para.default_input.asnumpy(), np.ones((1, 2, 3)))
+
+    # Call init_data() after default_input is set.
+    para = Parameter(initializer('ones', [1, 2, 3], mstype.float32), 'test2')
+    assert not isinstance(para.default_input, Tensor)
+    para.default_input = Tensor(np.zeros((1, 2, 3)))
+    assert np.array_equal(para.default_input.asnumpy(), np.zeros((1, 2, 3)))
+    para.init_data() # expect no effect.
+    assert np.array_equal(para.default_input.asnumpy(), np.zeros((1, 2, 3)))