Fix Tensor.from_numpy() returns wrong type

Tensor.from_numpy() should return mindspore.Tensor not _c_expression.Tensor.

mindspore/common/tensor.py

@@ -232,6 +232,11 @@ class Tensor(Tensor_):
             raise TypeError("virtual_flag must be bool.")
         self._virtual_flag = value
 
+    @staticmethod
+    def from_numpy(array):
+        """Convert numpy array to Tensor without copy data."""
+        return Tensor(Tensor_.from_numpy(array))
+
     def asnumpy(self):
         """Convert tensor to numpy array."""
         return Tensor_.asnumpy(self)

tests/ut/python/ir/test_tensor.py

@@ -480,6 +480,7 @@ def test_tensor_operation():
 def test_tensor_from_numpy():
     a = np.ones((2, 3))
     t = ms.Tensor.from_numpy(a)
+    assert isinstance(t, ms.Tensor)
     assert np.all(t.asnumpy() == 1)
     # 't' and 'a' share same data.
     a[1] = 2
@@ -489,3 +490,6 @@ def test_tensor_from_numpy():
     del a
     assert np.all(t.asnumpy()[0] == 1)
     assert np.all(t.asnumpy()[1] == 2)
+    with pytest.raises(TypeError):
+        # incorrect input.
+        t = ms.Tensor.from_numpy([1, 2, 3])