diff --git a/mindspore/core/base/core_ops.h b/mindspore/core/base/core_ops.h
index 696237d1a42..fe2f2e53d0a 100644
--- a/mindspore/core/base/core_ops.h
+++ b/mindspore/core/base/core_ops.h
@@ -50,6 +50,7 @@ constexpr auto kMul = "Mul";
 constexpr auto kRealDiv = "RealDiv";
 constexpr auto kAdd = "Add";
 constexpr auto kTile = "Tile";
+constexpr auto kBiasAddGrad = "BiasAddGrad";
 
 // Arrays
 constexpr auto kStack = "Stack";
@@ -274,7 +275,7 @@ inline const PrimitivePtr kPrimDepthwiseConv2dNativeBackpropInput =
   std::make_shared<Primitive>("DepthwiseConv2dNativeBackpropInput");
 inline const PrimitivePtr kPrimDetectionPostProcess = std::make_shared<Primitive>("DetectionPostProcess");
 inline const PrimitivePtr kPrimBiasAdd = std::make_shared<Primitive>("BiasAdd");
-inline const PrimitivePtr kPrimBiasAddGrad = std::make_shared<Primitive>("BiasAddGrad");
+inline const PrimitivePtr kPrimBiasAddGrad = std::make_shared<Primitive>(kBiasAddGrad);
 inline const PrimitivePtr kPrimBiasSubGrad = std::make_shared<Primitive>("BiasSubGrad");
 inline const PrimitivePtr kPrimBinaryCrossEntropy = std::make_shared<Primitive>("BinaryCrossEntropy");
 inline const PrimitivePtr kPrimBinaryCrossEntropyGrad = std::make_shared<Primitive>("BinaryCrossEntropyGrad");
diff --git a/mindspore/core/ops/grad/bias_add_grad.cc b/mindspore/core/ops/grad/bias_add_grad.cc
index 9fec587f180..5d8f269a14e 100644
--- a/mindspore/core/ops/grad/bias_add_grad.cc
+++ b/mindspore/core/ops/grad/bias_add_grad.cc
@@ -1,5 +1,5 @@
 /**
- * Copyright 2020 Huawei Technologies Co., Ltd
+ * Copyright 2020-2021 Huawei Technologies Co., Ltd
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -18,6 +18,7 @@
 #include <string>
 #include <algorithm>
 #include <memory>
+#include <map>
 #include <set>
 #include <vector>
 #include "ops/op_utils.h"
@@ -26,36 +27,55 @@
 
 namespace mindspore {
 namespace ops {
-void BiasAddGrad::Init(const Format format) { this->set_format(format); }
-
-void BiasAddGrad::set_format(const Format format) {
-  int64_t f = format;
-  AddAttr(kFormat, MakeValue(f));
+namespace {
+std::vector<int64_t> GetFormatShape(const int64_t &format, const std::vector<int64_t> &input_shape) {
+  std::vector<int64_t> output_shape;
+  if (format == NHWC) {
+    output_shape.push_back(input_shape.back());
+  } else {
+    output_shape.push_back(input_shape[1]);
+  }
+  return output_shape;
 }
-
-Format BiasAddGrad::get_format() const {
-  auto value_ptr = GetAttr(kFormat);
-  return Format(GetValue<int64_t>(value_ptr));
-}
-
-AbstractBasePtr BiasAddGradInfer(const abstract::AnalysisEnginePtr &, const PrimitivePtr &primitive,
-                                 const std::vector<AbstractBasePtr> &input_args) {
+abstract::ShapePtr InferShape(const PrimitivePtr &primitive, const std::vector<AbstractBasePtr> &input_args) {
   MS_EXCEPTION_IF_NULL(primitive);
   auto prim_name = primitive->name();
-  CheckAndConvertUtils::CheckInteger("bias_grad_infer", input_args.size(), kEqual, 1, prim_name);
-  MS_EXCEPTION_IF_NULL(input_args[0]);
-
-  // Infer shape
-  auto inshape = CheckAndConvertUtils::ConvertShapePtrToShapeMap(input_args[0]->BuildShape())[kShape];
-  for (size_t i = 0; i < inshape.size() - 1; i++) {
-    inshape[i] = 1;
+  CheckAndConvertUtils::CheckInteger("input numbers", input_args.size(), kEqual, 1, prim_name);
+  for (const auto &item : input_args) {
+    MS_EXCEPTION_IF_NULL(item);
   }
-
-  // Infer type
-  auto intype = input_args[0]->BuildType()->cast<TensorTypePtr>()->element();
-
-  return std::make_shared<abstract::AbstractTensor>(intype, inshape);
+  int64_t format = CheckAndConvertUtils::GetAndCheckFormat(primitive->GetAttr("format"));
+  auto shape_map = CheckAndConvertUtils::ConvertShapePtrToShapeMap(input_args[0]->BuildShape());
+  auto input_shape = shape_map[kShape];
+  auto min_shape = shape_map[kMinShape];
+  auto max_shape = shape_map[kMaxShape];
+  auto input_shape_ = GetFormatShape(format, input_shape);
+  if (min_shape.size() != 0 && max_shape.size() != 0) {
+    auto min_shape_ = GetFormatShape(format, min_shape);
+    auto max_shape_ = GetFormatShape(format, max_shape);
+    return std::make_shared<abstract::Shape>(input_shape_, min_shape_, max_shape_);
+  }
+  return std::make_shared<abstract::Shape>(input_shape_);
 }
-REGISTER_PRIMITIVE_C(kNameBiasAddGrad, BiasAddGrad);
+TypePtr InferType(const PrimitivePtr &prim, const std::vector<AbstractBasePtr> &input_args) {
+  MS_EXCEPTION_IF_NULL(prim);
+  auto prim_name = prim->name();
+  CheckAndConvertUtils::CheckInteger("BiasAddGrad infer", input_args.size(), kEqual, 1, prim_name);
+  MS_EXCEPTION_IF_NULL(input_args[0]);
+  auto x_type_map = input_args[0]->BuildType();
+  MS_EXCEPTION_IF_NULL(x_type_map);
+  auto x_type = x_type_map->cast<TensorTypePtr>();
+  MS_EXCEPTION_IF_NULL(x_type);
+  std::set<TypePtr> valid_x_type = {kTensorType};
+  return CheckAndConvertUtils::CheckTensorTypeValid("input_x", x_type, valid_x_type, prim_name);
+}
+}  // namespace
+AbstractBasePtr BiasAddGradInfer(const abstract::AnalysisEnginePtr &, const PrimitivePtr &primitive,
+                                 const std::vector<AbstractBasePtr> &input_args) {
+  return std::make_shared<abstract::AbstractTensor>(InferType(primitive, input_args),
+                                                    InferShape(primitive, input_args)->shape());
+}
+REGISTER_PRIMITIVE_EVAL_IMPL(BiasAddGrad, prim::kPrimBiasAddGrad, BiasAddGradInfer, nullptr, true);
+
 }  // namespace ops
 }  // namespace mindspore
diff --git a/mindspore/core/ops/grad/bias_add_grad.h b/mindspore/core/ops/grad/bias_add_grad.h
index 7399a77ccad..b1df50986de 100644
--- a/mindspore/core/ops/grad/bias_add_grad.h
+++ b/mindspore/core/ops/grad/bias_add_grad.h
@@ -1,5 +1,5 @@
 /**
- * Copyright 2020 Huawei Technologies Co., Ltd
+ * Copyright 2020-2021 Huawei Technologies Co., Ltd
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -26,15 +26,13 @@
 
 namespace mindspore {
 namespace ops {
-constexpr auto kNameBiasAddGrad = "BiasAddGrad";
+constexpr auto kNameBiasAddGrad = prim::kBiasAddGrad;
 class BiasAddGrad : public PrimitiveC {
  public:
-  BiasAddGrad() : PrimitiveC(kNameBiasAddGrad) {}
+  BiasAddGrad() : PrimitiveC(prim::kPrimBiasAddGrad->name()) { InitIOName({"x"}, {"output"}); }
   ~BiasAddGrad() = default;
   MS_DECLARE_PARENT(BiasAddGrad, PrimitiveC);
-  void Init(const Format format);
-  void set_format(const Format format);
-  Format get_format() const;
+  void Init() {}
 };
 AbstractBasePtr BiasAddGradInfer(const abstract::AnalysisEnginePtr &, const PrimitivePtr &primitive,
                                  const std::vector<AbstractBasePtr> &input_args);
diff --git a/mindspore/ops/operations/_grad_ops.py b/mindspore/ops/operations/_grad_ops.py
index bfb52fe1cfd..0bd64b50824 100644
--- a/mindspore/ops/operations/_grad_ops.py
+++ b/mindspore/ops/operations/_grad_ops.py
@@ -224,7 +224,7 @@ class SyncBatchNormGrad(PrimitiveWithInfer):
         return (x_type, scale_type, scale_type)
 
 
-class BiasAddGrad(PrimitiveWithInfer):
+class BiasAddGrad(Primitive):
     """Computes gradients of BiasAdd."""
 
     @prim_attr_register
@@ -237,13 +237,6 @@ class BiasAddGrad(PrimitiveWithInfer):
             self.format = "NCHW"
         self.add_prim_attr('data_format', self.format)
 
-    def infer_shape(self, d_output):
-        channel = d_output[-1] if self.format == "NHWC" else d_output[1]
-        return (channel,)
-
-    def infer_dtype(self, dout_dtype):
-        return dout_dtype
-
 
 class KLDivLossGrad(PrimitiveWithInfer):
     """Computes gradients for `KLDivLoss` operation."""