!15072 numpy-native fix support of non tensor input for isin and reduce ufuncs

From: @jachua
Reviewed-by: @c_34,@guoqi1024
Signed-off-by: @c_34,@guoqi1024

mindspore/numpy/array_ops.py

@@ -669,7 +669,7 @@ def where(condition, x=None, y=None):
 
     Args:
         condition (Tensor): where True, yield `x`, otherwise yield `y`.
-        x (Tensor)
+        x (Tensor): Values from which to choose.
         y (Tensor): Values from which to choose. `x`, `y` and `condition` need
             to be broadcastable to some shape.
 

mindspore/numpy/logic_ops.py

@@ -578,9 +578,9 @@ def isin(element, test_elements, invert=False):
         not rely on the uniqueness of the input arrays.
 
     Args:
-        element (array_like): Input array.
-        test_elements (array_like): The values against which to test each value of
-            `element`.
+        element (Union[int, float, bool, list, tuple, Tensor]): Input array.
+        test_elements (Union[int, float, bool, list, tuple, Tensor]): The values against
+            which to test each value of `element`.
         invert (boolean, optional): If True, the values in the returned array are
             inverted, as if calculating `element` not in `test_elements`. Default is False.
 
@@ -603,6 +603,7 @@ def isin(element, test_elements, invert=False):
         [[ True False]
         [False  True]]
     """
+    element = _to_tensor(element)
     res = in1d(element, test_elements, invert=invert)
     return F.reshape(res, F.shape(element))
 

mindspore/numpy/math_ops.py

@@ -378,6 +378,7 @@ def divide(x1, x2, dtype=None):
         [0.33333334 0.5       ]
         [0.33333334 0.5       ]]
     """
+    x1, x2 = _to_tensor(x1, x2)
     if not _check_is_float(F.dtype(x1)) and not _check_is_float(F.dtype(x2)):
         x1 = F.cast(x1, mstype.float32)
         x2 = F.cast(x2, mstype.float32)
@@ -2427,7 +2428,7 @@ def _reduce(a, reduce_fn, cmp_fn=None, axis=None, keepdims=False, initial=None,
     Applies comparison based on cmp_fn and reduction based on reduce_fn.
     If cmp_fn is None, only reduction is performed.
     """
-    _check_input_tensor(a)
+    a = _to_tensor(a)
 
     shape = F.shape(a)
     ndim = F.rank(a)
@@ -2458,8 +2459,6 @@ def _reduce(a, reduce_fn, cmp_fn=None, axis=None, keepdims=False, initial=None,
     if initial is not None:
         initial = full(shape, initial, dtype)
         a = cmp_fn(a, initial)
-    if not axes:
-        return a.astype(dtype)
     if isinstance(where, Tensor):
         if initial is None:
             return _raise_value_error('initial value must be provided for where masks')
@@ -2580,6 +2579,8 @@ def nanmean(a, axis=None, dtype=None, keepdims=False):
         >>> print(output)
         [1.  3.5]
     """
+    if dtype is None:
+        dtype = mstype.float32
     a = _to_tensor(a)
     axis = _check_axis_valid(axis, F.rank(a))
     sum_a = nansum(a, axis=axis, dtype=dtype, keepdims=keepdims)
@@ -2592,7 +2593,7 @@ def _nanvar(a, axis, ddof=0, keepdims=False):
     pow_a = F.tensor_pow(F.tensor_sub(a, mean_a), 2)
     sum_a = _reduce_nansum(pow_a, axis, keepdims)
     count = _count_nonnan(a, axis, keepdims)
-    return F.tensor_div(sum_a, F.tensor_sub(count, ddof))
+    return divide(sum_a, F.tensor_sub(count, ddof))
 
 
 def nanvar(a, axis=None, dtype=None, ddof=0, keepdims=False):
@@ -2633,16 +2634,18 @@ def nanvar(a, axis=None, dtype=None, ddof=0, keepdims=False):
     Examples:
         >>> import mindspore.numpy as np
         >>> a = np.array([[1, np.nan], [3, 4]])
-        >>> output = np.nanstd(a)
+        >>> output = np.nanvar(a)
         >>> print(output)
-        1.2472192
-        >>> output = np.nanstd(a, axis=0)
+        1.5555557
+        >>> output = np.nanvar(a, axis=0)
         >>> print(output)
         [1. 0.]
-        >>> output = np.nanstd(a, axis=1)
+        >>> output = np.nanvar(a, axis=1)
         >>> print(output)
-        [0.  0.5]
+        [0.   0.25]
     """
+    if dtype is None:
+        dtype = mstype.float32
     return _reduce(a, functools.partial(_nanvar, ddof=ddof, keepdims=keepdims), axis=axis,
                    keepdims=keepdims, dtype=dtype)
 
@@ -2686,16 +2689,18 @@ def nanstd(a, axis=None, dtype=None, ddof=0, keepdims=False):
     Examples:
         >>> import mindspore.numpy as np
         >>> a = np.array([[1, np.nan], [3, 4]])
-        >>> output = np.nanvar(a)
+        >>> output = np.nanstd(a)
         >>> print(output)
-        1.5555557
-        >>> output = np.nanvar(a, axis=0)
+        1.2472192
+        >>> output = np.nanstd(a, axis=0)
         >>> print(output)
         [1. 0.]
-        >>> output = np.nanvar(a, axis=1)
+        >>> output = np.nanstd(a, axis=1)
         >>> print(output)
-        [0.   0.25]
+        [0.  0.5]
     """
+    if dtype is None:
+        dtype = mstype.float32
     return _reduce(a, lambda a, axis: F.sqrt(_nanvar(a, axis, ddof=ddof, keepdims=keepdims)),
                    axis=axis, keepdims=keepdims, dtype=dtype)
 

tests/st/numpy_native/test_math_ops.py

@@ -370,6 +370,7 @@ def test_nanstd():
     arr2[0][4][3][0][2] = onp.nan
     run_single_test(mnp_nanstd, onp_nanstd, arr1, error=1e-5)
     run_single_test(mnp_nanstd, onp_nanstd, arr2, error=1e-5)
+    match_res(mnp.nanstd, onp.nanstd, rand_int())
 
 
 def mnp_var(x):
@@ -436,6 +437,7 @@ def test_nanvar():
     arr2[0][4][3][0][2] = onp.nan
     run_single_test(mnp_nanvar, onp_nanvar, arr1, error=1e-5)
     run_single_test(mnp_nanvar, onp_nanvar, arr2, error=1e-5)
+    match_res(mnp.nanvar, onp.nanvar, rand_int())
 
 
 def mnp_average(x):
@@ -993,6 +995,7 @@ def test_amax():
     a = rand_int(2, 3, 4, 5).astype('float32')
     mask = rand_bool(2, 3, 4, 5)
     run_multi_test(mnp_amax, onp_amax, (a, mask))
+    match_res(mnp.amax, onp.amax, rand_int())
 
 
 def mnp_amin(x, mask):
@@ -1029,6 +1032,7 @@ def test_amin():
     a = rand_int(2, 3, 4, 5).astype('float32')
     mask = rand_bool(2, 3, 4, 5)
     run_multi_test(mnp_amin, onp_amin, (a, mask))
+    match_res(mnp.amin, onp.amin, rand_int())
 
 
 def mnp_hypot(x1, x2):
@@ -1843,6 +1847,7 @@ def test_nansum():
     x[1][0][2][4] = onp.nan
     x[1][1][1][1] = onp.nan
     run_multi_test(mnp_nansum, onp_nansum, (x,))
+    match_res(mnp.nansum, onp.nansum, rand_int())
 
 
 def mnp_nanmean(x):
@@ -1875,6 +1880,7 @@ def test_nanmean():
     x[1][0][2][4] = onp.nan
     x[1][1][1][1] = onp.nan
     run_multi_test(mnp_nanmean, onp_nanmean, (x,))
+    match_res(mnp.nanmean, onp.nanmean, rand_int())
 
 
 def mnp_mean(*arrs):