fix(mge/tensor): implement abtract method to fix lint errors

GitOrigin-RevId: d53f2eac6a

imperative/python/megengine/tensor.py

@@ -6,9 +6,7 @@
 # Unless required by applicable law or agreed to in writing,
 # software distributed under the License is distributed on an
 # "AS IS" BASIS, WITHOUT ARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-
-
-import collections
+from typing import Union
 
 import numpy as np
 
@@ -53,7 +51,7 @@ class Tensor(_Tensor, ArrayMethodMixin):
         return obj
 
     @property
-    def shape(self):
+    def shape(self) -> Union[tuple, "Tensor"]:
         shape = super().shape
         if shape == () or not use_symbolic_shape():
             return shape
@@ -63,6 +61,16 @@ class Tensor(_Tensor, ArrayMethodMixin):
     def _tuple_shape(self):
         return super().shape
 
+    @property
+    def dtype(self) -> np.dtype:
+        return super().dtype
+
+    def numpy(self) -> np.ndarray:
+        return super().numpy()
+
+    def _reset(self, other):
+        super()._reset(other)
+
     def __repr__(self):
         piece = "Tensor("
         with np.printoptions(precision=4, suppress=True):