!8505 adapt implicit conversion to the change that python int resolved as int64 instead of int32

From: @zhangbuxue
Reviewed-by: @chenfei52,@zh_qh
Signed-off-by: @zh_qh

mindspore/ccsrc/frontend/operator/composite/do_signature.cc

@@ -96,7 +96,7 @@ TypeId GetMaxTypeId(const std::vector<TypePtr> &input_types, std::vector<size_t>
   TypeId max_type_id = kTypeUnknown;
   size_t max_type_number = 0;
   bool has_int8 = false;
-  bool has_scalar_int32 = false;
+  bool has_scalar_int64 = false;
   bool has_scalar_float32 = false;
   for (const auto &index : indices) {
     TypeId arg_type_id = kTypeUnknown;
@@ -105,8 +105,8 @@ TypeId GetMaxTypeId(const std::vector<TypePtr> &input_types, std::vector<size_t>
       continue;
     }
     if (arg_type != kObjectTypeTensorType) {
-      if (arg_type_id == kNumberTypeInt32) {
-        has_scalar_int32 = true;
+      if (arg_type_id == kNumberTypeInt64) {
+        has_scalar_int64 = true;
       } else if (arg_type_id == kNumberTypeFloat32) {
         has_scalar_float32 = true;
       }
@@ -135,8 +135,8 @@ TypeId GetMaxTypeId(const std::vector<TypePtr> &input_types, std::vector<size_t>
   // if so, it means that max is bool tensor, use scalar type instead.
   // for example: Tensor([True, True]) * 2, expect result is Tensor([2, 2])
   if (max_type_id == kNumberTypeBool) {
-    if (has_scalar_int32) {
-      max_type_id = kNumberTypeInt32;
+    if (has_scalar_int64) {
+      max_type_id = kNumberTypeInt64;
     }
     if (has_scalar_float32) {
       max_type_id = kNumberTypeFloat32;

mindspore/ccsrc/pipeline/pynative/pynative_execute.cc

@@ -181,15 +181,15 @@ std::map<SignatureEnumDType, TypeId> GetDstType(const py::tuple &py_args,
     }
     size_t priority = 0;
     TypeId max_type = TypeId::kTypeUnknown;
-    bool has_float = false;
-    bool has_int = false;
-    bool has_int8 = false;
+    bool has_scalar_float32 = false;
+    bool has_scalar_int64 = false;
+    bool has_tensor_int8 = false;
     for (size_t index : indexes) {
-      if (!has_float && py::isinstance<py::float_>(py_args[index])) {
-        has_float = true;
+      if (!has_scalar_float32 && py::isinstance<py::float_>(py_args[index])) {
+        has_scalar_float32 = true;
       }
-      if (!has_int && !py::isinstance<py::bool_>(py_args[index]) && py::isinstance<py::int_>(py_args[index])) {
-        has_int = true;
+      if (!has_scalar_int64 && !py::isinstance<py::bool_>(py_args[index]) && py::isinstance<py::int_>(py_args[index])) {
+        has_scalar_int64 = true;
       }
 
       auto obj = py_args[index];
@@ -201,7 +201,7 @@ std::map<SignatureEnumDType, TypeId> GetDstType(const py::tuple &py_args,
           continue;
         }
         if (arg_type_id == kNumberTypeInt8) {
-          has_int8 = true;
+          has_tensor_int8 = true;
         }
         if (type_priority->second > priority) {
           max_type = type_priority->first;
@@ -210,18 +210,18 @@ std::map<SignatureEnumDType, TypeId> GetDstType(const py::tuple &py_args,
       }
     }
     if (max_type == TypeId::kNumberTypeBool) {
-      if (has_int) {
+      if (has_scalar_int64) {
         max_type = TypeId::kNumberTypeInt64;
       }
-      if (has_float) {
+      if (has_scalar_float32) {
         max_type = TypeId::kNumberTypeFloat32;
       }
     }
     if (max_type != TypeId::kNumberTypeFloat16 && max_type != TypeId::kNumberTypeFloat32 &&
-        max_type != TypeId::kNumberTypeFloat64 && max_type != TypeId::kTypeUnknown && has_float) {
+        max_type != TypeId::kNumberTypeFloat64 && max_type != TypeId::kTypeUnknown && has_scalar_float32) {
       max_type = TypeId::kNumberTypeFloat32;
     }
-    if (max_type == TypeId::kNumberTypeUInt8 && has_int8) {
+    if (max_type == TypeId::kNumberTypeUInt8 && has_tensor_int8) {
       max_type = TypeId::kNumberTypeInt16;
     }
     (void)dst_type.emplace(std::make_pair(type, max_type));

mindspore/ops/operations/nn_ops.py

@@ -1780,8 +1780,7 @@ class TopK(PrimitiveWithInfer):
         >>> input_x = Tensor([1, 2, 3, 4, 5], mindspore.float16)
         >>> k = 3
         >>> values, indices = topk(input_x, k)
-        >>> assert values == Tensor(np.array([5, 4, 3]), mstype.float16).all()
-        >>> assert indices == Tensor(np.array([4, 3, 2]), mstype.int32).all()
+        ([5.0, 4.0, 3.0], [4, 3, 2])
     """
 
     @prim_attr_register