fix pad ops doc and nulltpr

docs/api/api_python/ops/mindspore.ops.PadV3.rst

@@ -24,8 +24,8 @@ mindspore.ops.PadV3
         - **ValueError** - `mode` 是"constant"的同时 `paddings` 元素个数不是偶数。
         - **ValueError** - `mode` 是"constant"的同时 `paddings` 元素个数大于输入维度乘以2。
         - **ValueError** - `mode` 是"edge"或"reflect"的同时 `paddings` 元素个数不是2、4或6。
-        - **ValueError** - `mode` 是"edge"或"reflect"， `x` 的维度是3， `paddings` 元素个数是2。
+        - **ValueError** - `mode` 是"edge"或"reflect"， `x` 的维度是3， `paddings` 元素个数不是2。
-        - **ValueError** - `mode` 是"edge"或"reflect"， `x` 的维度是4， `paddings` 元素个数是4。
+        - **ValueError** - `mode` 是"edge"或"reflect"， `x` 的维度是4， `paddings` 元素个数不是4。
         - **ValueError** - `mode` 是"edge"或"reflect"的同时 `x` 的维度小于3。
         - **ValueError** - `mode` 是"edge"的同时 `x` 的维度大于5。
         - **ValueError** - `mode` 是"reflect"的同时 `x` 的维度大于4。

mindspore/core/ops/mirror_pad.cc

@@ -51,43 +51,13 @@ void CheckPaddingParam(const std::vector<int64_t> &paddings_shape, const std::ve
   return;
 }
 
-MIND_API_OPERATOR_IMPL(MirrorPad, BaseOperator);
+void CheckPaddingValue(const std::vector<std::pair<int64_t, int64_t>> &paddings_attr,
-class MirrorPadInfer : public abstract::OpInferBase {
+                       const std::vector<int64_t> &x_shape, const std::string &mode, const std::string &prim_name) {
- public:
-  BaseShapePtr InferShape(const PrimitivePtr &primitive,
-                          const std::vector<AbstractBasePtr> &input_args) const override {
-    MS_EXCEPTION_IF_NULL(primitive);
-    auto prim_name = primitive->name();
-    auto input_x_shape_ptr = input_args[0]->BuildShape();
-    MS_EXCEPTION_IF_NULL(input_x_shape_ptr);
-    auto input_x_shape = input_x_shape_ptr->cast<abstract::ShapePtr>();
-    // Dynamic rank process.
-    if (IsDynamicRank(input_x_shape->shape())) {
-      return std::make_shared<abstract::Shape>(ShapeVector{abstract::Shape::kShapeRankAny});
-    }
-    auto paddings = input_args[1]->BuildValue();
-    MS_EXCEPTION_IF_NULL(paddings);
-    auto x_shape = CheckAndConvertUtils::ConvertShapePtrToShapeMap(input_args[0]->BuildShape())[kShape];
-    // if shape of x is determined and padding value is unknown, return a all -1 shape
-    if (paddings->isa<AnyValue>() || paddings->isa<None>()) {
-      return std::make_shared<abstract::Shape>(ShapeVector(x_shape.size(), abstract::Shape::kShapeDimAny));
-    }
-    auto paddings_arg = CheckAndConvertUtils::CheckTensorIntValue(kPaddings, paddings, prim_name);
-    std::vector<std::pair<int64_t, int64_t>> paddings_attr;
-
-    auto paddings_shape = CheckAndConvertUtils::ConvertShapePtrToShapeMap(input_args[1]->BuildShape())[kShape];
-    auto mode = GetValue<std::string>(primitive->GetAttr(kMode));
-    CheckPaddingParam(paddings_shape, x_shape, prim_name);
-    for (size_t i = 0; i < paddings_arg.size(); i = i + static_cast<size_t>(kPaddingsSecondDimSize)) {
-      paddings_attr.push_back(std::make_pair(paddings_arg[i], paddings_arg[i + 1]));
-    }
-    (void)CheckAndConvertUtils::CheckInteger(kPaddingsSize, SizeToLong(paddings_attr.size()), kEqual,
-                                             SizeToLong(x_shape.size()), prim_name);
   int64_t size = static_cast<int64_t>(x_shape.size());
   if (size < 0 || size > MAX_PADDINGS) {
     MS_EXCEPTION(ValueError) << "For '" << prim_name
-                               << "', the dimension of input only supports less than or equal to 5 dims, but got "
+                             << "', the dimension of input only supports less than or equal to 5 dims, but got " << size
-                               << size << " dims";
+                             << " dims";
   }
   for (int64_t i = 0; i < size; i++) {
     if (x_shape[i] == abstract::Shape::kShapeDimAny) {
@@ -116,6 +86,44 @@ class MirrorPadInfer : public abstract::OpInferBase {
       }
     }
   }
+}
+
+MIND_API_OPERATOR_IMPL(MirrorPad, BaseOperator);
+class MirrorPadInfer : public abstract::OpInferBase {
+ public:
+  BaseShapePtr InferShape(const PrimitivePtr &primitive,
+                          const std::vector<AbstractBasePtr> &input_args) const override {
+    MS_EXCEPTION_IF_NULL(primitive);
+    auto prim_name = primitive->name();
+    for (const auto &item : input_args) {
+      MS_EXCEPTION_IF_NULL(item);
+    }
+    auto input_x_shape_ptr = input_args[0]->BuildShape();
+    MS_EXCEPTION_IF_NULL(input_x_shape_ptr);
+    auto input_x_shape = input_x_shape_ptr->cast<abstract::ShapePtr>();
+    // Dynamic rank process.
+    if (IsDynamicRank(input_x_shape->shape())) {
+      return std::make_shared<abstract::Shape>(ShapeVector{abstract::Shape::kShapeRankAny});
+    }
+    auto paddings = input_args[1]->BuildValue();
+    MS_EXCEPTION_IF_NULL(paddings);
+    auto x_shape = CheckAndConvertUtils::ConvertShapePtrToShapeMap(input_args[0]->BuildShape())[kShape];
+    // if shape of x is determined and padding value is unknown, return a all -1 shape
+    if (paddings->isa<AnyValue>() || paddings->isa<None>()) {
+      return std::make_shared<abstract::Shape>(ShapeVector(x_shape.size(), abstract::Shape::kShapeDimAny));
+    }
+    auto paddings_arg = CheckAndConvertUtils::CheckTensorIntValue(kPaddings, paddings, prim_name);
+    std::vector<std::pair<int64_t, int64_t>> paddings_attr;
+
+    auto paddings_shape = CheckAndConvertUtils::ConvertShapePtrToShapeMap(input_args[1]->BuildShape())[kShape];
+    auto mode = GetValue<std::string>(primitive->GetAttr(kMode));
+    CheckPaddingParam(paddings_shape, x_shape, prim_name);
+    for (size_t i = 0; i < paddings_arg.size(); i = i + static_cast<size_t>(kPaddingsSecondDimSize)) {
+      paddings_attr.push_back(std::make_pair(paddings_arg[i], paddings_arg[i + 1]));
+    }
+    (void)CheckAndConvertUtils::CheckInteger(kPaddingsSize, SizeToLong(paddings_attr.size()), kEqual,
+                                             SizeToLong(x_shape.size()), prim_name);
+    CheckPaddingValue(paddings_attr, x_shape, mode, prim_name);
     std::vector<int64_t> out_shape;
     for (size_t i = 0; i < x_shape.size(); i++) {
       // In dynamic situation , if input axis is dynamic, output axis is dynamic too.

mindspore/core/ops/padding.cc

@@ -31,6 +31,9 @@ namespace {
 TypePtr PaddingInferType(const PrimitivePtr &primitive, const std::vector<AbstractBasePtr> &input_args) {
   MS_EXCEPTION_IF_NULL(primitive);
   auto name = primitive->name();
+  for (const auto &item : input_args) {
+    MS_EXCEPTION_IF_NULL(item);
+  }
   auto context = MsContext::GetInstance();
   MS_EXCEPTION_IF_NULL(context);
   bool is_gpu = (context->get_param<std::string>(MS_CTX_DEVICE_TARGET) == kGPUDevice);

mindspore/python/mindspore/ops/operations/nn_ops.py

@@ -4298,9 +4298,9 @@ class PadV3(Primitive):
         ValueError: If `mode` is "constant", the element's number of `paddings` large than input dim * 2.
         ValueError: If `mode` is "edge" or "reflect", the element's number of `paddings` is not 2, 4 or 6.
         ValueError: If `mode` is "edge" or "reflect", `x` dims equals 3,
-            the element's number of `paddings` is 2.
+            the element's number of `paddings` is not 2.
         ValueError: If `mode` is "edge" or "reflect", `x` dims equals 4,
-            the element's number of `paddings` is 4.
+            the element's number of `paddings` is not 4.
         ValueError: If `mode` is "edge" or "reflect", `x` dims smaller than 3.
         ValueError: If `mode` is "edge", x dims bigger than 5.
         ValueError: If `mode` is "reflect", x dims bigger than 4.