diff --git a/mindspore/ccsrc/backend/kernel_compiler/cpu/sparse_to_dense_cpu_kernal.cc b/mindspore/ccsrc/backend/kernel_compiler/cpu/sparse_to_dense_cpu_kernal.cc
new file mode 100644
index 00000000000..a48f4c94189
--- /dev/null
+++ b/mindspore/ccsrc/backend/kernel_compiler/cpu/sparse_to_dense_cpu_kernal.cc
@@ -0,0 +1,98 @@
+/**
+ * Copyright 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.
+ * You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include "backend/kernel_compiler/cpu/sparse_to_dense_cpu_kernal.h"
+#include <algorithm>
+#include "runtime/device/cpu/cpu_device_address.h"
+
+namespace mindspore {
+namespace kernel {
+template <typename I, typename T>
+void SparseToDenseCPUKernel<I, T>::InitKernel(const CNodePtr &kernel_node) {
+  CheckParam(kernel_node);
+  indices_shape_ = AnfAlgo::GetPrevNodeOutputInferShape(kernel_node, 0);
+  values_shape_ = AnfAlgo::GetPrevNodeOutputInferShape(kernel_node, 1);
+  dense_shape_shape_ = AnfAlgo::GetPrevNodeOutputInferShape(kernel_node, 2);
+  output_shape_ = AnfAlgo::GetOutputInferShape(kernel_node, 0);
+  if (!indices_shape_.size() || !values_shape_.size() || !output_shape_.size()) {
+    MS_LOG(EXCEPTION) << "Input NULL";
+  }
+  if (indices_shape_.size() > 2 || indices_shape_[0] != values_shape_[0]) {
+    MS_LOG(EXCEPTION) << "Input Error";
+  }
+}
+
+size_t DenseGetTensorLen(const std::vector<size_t> &shape) {
+  size_t len = 1;
+  for (size_t i = 0; i < shape.size(); i++) {
+    len *= shape[i];
+  }
+  return len;
+}
+
+template <typename I, typename T>
+bool SparseToDenseCPUKernel<I, T>::Launch(const std::vector<kernel::AddressPtr> &inputs,
+                                          const std::vector<kernel::AddressPtr> & /*workspace*/,
+                                          const std::vector<kernel::AddressPtr> &outputs) {
+  auto indices_addr = reinterpret_cast<I *>(inputs[0]->addr);
+  auto values_addr = reinterpret_cast<T *>(inputs[1]->addr);
+  auto output_addr = reinterpret_cast<T *>(outputs[0]->addr);
+
+  size_t output_len = DenseGetTensorLen(output_shape_);
+  memset(output_addr, 0, output_len * sizeof(T));
+  std::vector<size_t> cargo(output_shape_.size(), 0);
+
+  size_t i = output_shape_.size() - 1;
+  switch (indices_shape_.size()) {
+    case 1:
+      for (i = 0; i < indices_shape_[0]; i++) {
+        output_addr[indices_addr[i]] = values_addr[i];
+      }
+      break;
+
+    case 2:
+      cargo[i] = 1;
+      for (; i >= 1; i--) {
+        cargo[i - 1] = cargo[i] * output_shape_[i];
+      }
+      for (i = 0; i < indices_shape_[0]; i++) {
+        size_t out_index = 0;
+        for (size_t j = 0; j < indices_shape_[1]; j++) {
+          out_index += (*(indices_addr + i * indices_shape_[1] + j)) * cargo[j];
+        }
+        output_addr[out_index] = values_addr[i];
+      }
+      break;
+
+    default:
+      break;
+  }
+  return true;
+}
+
+template <typename I, typename T>
+void SparseToDenseCPUKernel<I, T>::CheckParam(const CNodePtr &kernel_node) {
+  size_t input_num = AnfAlgo::GetInputTensorNum(kernel_node);
+  if (input_num != 3) {
+    MS_LOG(EXCEPTION) << "Input number is " << input_num << ", but SparseToDenseCPUKernel needs 3 input.";
+  }
+  size_t output_num = AnfAlgo::GetOutputTensorNum(kernel_node);
+  if (output_num != 1) {
+    MS_LOG(EXCEPTION) << "Output number is " << output_num << ", but SparseToDenseCPUKernel needs 1 output.";
+  }
+}
+}  // namespace kernel
+}  // namespace mindspore
diff --git a/mindspore/ccsrc/backend/kernel_compiler/cpu/sparse_to_dense_cpu_kernal.h b/mindspore/ccsrc/backend/kernel_compiler/cpu/sparse_to_dense_cpu_kernal.h
new file mode 100644
index 00000000000..d15820549a1
--- /dev/null
+++ b/mindspore/ccsrc/backend/kernel_compiler/cpu/sparse_to_dense_cpu_kernal.h
@@ -0,0 +1,207 @@
+/**
+ * Copyright 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.
+ * You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef MINDSPORE_CCSRC_BACKEND_KERNEL_COMPILER_CPU_SPARSETODENSE_CPU_KERNEL_H_
+#define MINDSPORE_CCSRC_BACKEND_KERNEL_COMPILER_CPU_SPARSETODENSE_CPU_KERNEL_H_
+#include <memory>
+#include <unordered_map>
+#include <vector>
+#include "backend/kernel_compiler/cpu/cpu_kernel.h"
+#include "backend/kernel_compiler/cpu/cpu_kernel_factory.h"
+
+namespace mindspore {
+namespace kernel {
+template <typename I, typename T>
+class SparseToDenseCPUKernel : public CPUKernel {
+ public:
+  SparseToDenseCPUKernel() = default;
+  ~SparseToDenseCPUKernel() override = default;
+
+  void InitKernel(const CNodePtr &kernel_node) override;
+
+  bool Launch(const std::vector<AddressPtr> &inputs, const std::vector<AddressPtr> &workspace,
+              const std::vector<AddressPtr> &outputs) override;
+
+ private:
+  void CheckParam(const CNodePtr &kernel_node);
+  std::vector<size_t> indices_shape_;
+  std::vector<size_t> values_shape_;
+  std::vector<size_t> dense_shape_shape_;
+  std::vector<size_t> output_shape_;
+};
+
+MS_REG_CPU_KERNEL_T_S(SparseToDense,
+                      KernelAttr()
+                        .AddInputAttr(kNumberTypeInt32)
+                        .AddInputAttr(kNumberTypeInt32)
+                        .AddInputAttr(kNumberTypeInt32)
+                        .AddOutputAttr(kNumberTypeInt32),
+                      SparseToDenseCPUKernel, int32_t, int32_t);
+
+MS_REG_CPU_KERNEL_T_S(SparseToDense,
+                      KernelAttr()
+                        .AddInputAttr(kNumberTypeInt32)
+                        .AddInputAttr(kNumberTypeInt8)
+                        .AddInputAttr(kNumberTypeInt32)
+                        .AddOutputAttr(kNumberTypeInt8),
+                      SparseToDenseCPUKernel, int32_t, int8_t);
+
+MS_REG_CPU_KERNEL_T_S(SparseToDense,
+                      KernelAttr()
+                        .AddInputAttr(kNumberTypeInt32)
+                        .AddInputAttr(kNumberTypeUInt8)
+                        .AddInputAttr(kNumberTypeInt32)
+                        .AddOutputAttr(kNumberTypeUInt8),
+                      SparseToDenseCPUKernel, int32_t, uint8_t);
+
+MS_REG_CPU_KERNEL_T_S(SparseToDense,
+                      KernelAttr()
+                        .AddInputAttr(kNumberTypeInt32)
+                        .AddInputAttr(kNumberTypeInt16)
+                        .AddInputAttr(kNumberTypeInt32)
+                        .AddOutputAttr(kNumberTypeInt16),
+                      SparseToDenseCPUKernel, int32_t, int16_t);
+
+MS_REG_CPU_KERNEL_T_S(SparseToDense,
+                      KernelAttr()
+                        .AddInputAttr(kNumberTypeInt32)
+                        .AddInputAttr(kNumberTypeUInt16)
+                        .AddInputAttr(kNumberTypeInt32)
+                        .AddOutputAttr(kNumberTypeUInt16),
+                      SparseToDenseCPUKernel, int32_t, uint16_t);
+
+MS_REG_CPU_KERNEL_T_S(SparseToDense,
+                      KernelAttr()
+                        .AddInputAttr(kNumberTypeInt32)
+                        .AddInputAttr(kNumberTypeInt64)
+                        .AddInputAttr(kNumberTypeInt32)
+                        .AddOutputAttr(kNumberTypeInt64),
+                      SparseToDenseCPUKernel, int32_t, int64_t);
+
+MS_REG_CPU_KERNEL_T_S(SparseToDense,
+                      KernelAttr()
+                        .AddInputAttr(kNumberTypeInt32)
+                        .AddInputAttr(kNumberTypeFloat16)
+                        .AddInputAttr(kNumberTypeInt32)
+                        .AddOutputAttr(kNumberTypeFloat16),
+                      SparseToDenseCPUKernel, int32_t, float16);
+
+MS_REG_CPU_KERNEL_T_S(SparseToDense,
+                      KernelAttr()
+                        .AddInputAttr(kNumberTypeInt32)
+                        .AddInputAttr(kNumberTypeFloat)
+                        .AddInputAttr(kNumberTypeInt32)
+                        .AddOutputAttr(kNumberTypeFloat),
+                      SparseToDenseCPUKernel, int32_t, float);
+
+MS_REG_CPU_KERNEL_T_S(SparseToDense,
+                      KernelAttr()
+                        .AddInputAttr(kNumberTypeInt32)
+                        .AddInputAttr(kNumberTypeFloat64)
+                        .AddInputAttr(kNumberTypeInt32)
+                        .AddOutputAttr(kNumberTypeFloat64),
+                      SparseToDenseCPUKernel, int32_t, double);
+
+MS_REG_CPU_KERNEL_T_S(SparseToDense,
+                      KernelAttr()
+                        .AddInputAttr(kNumberTypeInt32)
+                        .AddInputAttr(kNumberTypeBool)
+                        .AddInputAttr(kNumberTypeInt32)
+                        .AddOutputAttr(kNumberTypeBool),
+                      SparseToDenseCPUKernel, int32_t, bool);
+
+MS_REG_CPU_KERNEL_T_S(SparseToDense,
+                      KernelAttr()
+                        .AddInputAttr(kNumberTypeInt64)
+                        .AddInputAttr(kNumberTypeInt32)
+                        .AddInputAttr(kNumberTypeInt64)
+                        .AddOutputAttr(kNumberTypeInt32),
+                      SparseToDenseCPUKernel, int64_t, int32_t);
+
+MS_REG_CPU_KERNEL_T_S(SparseToDense,
+                      KernelAttr()
+                        .AddInputAttr(kNumberTypeInt64)
+                        .AddInputAttr(kNumberTypeInt64)
+                        .AddInputAttr(kNumberTypeInt64)
+                        .AddOutputAttr(kNumberTypeInt64),
+                      SparseToDenseCPUKernel, int64_t, int64_t);
+
+MS_REG_CPU_KERNEL_T_S(SparseToDense,
+                      KernelAttr()
+                        .AddInputAttr(kNumberTypeInt64)
+                        .AddInputAttr(kNumberTypeInt8)
+                        .AddInputAttr(kNumberTypeInt64)
+                        .AddOutputAttr(kNumberTypeInt8),
+                      SparseToDenseCPUKernel, int64_t, int8_t);
+
+MS_REG_CPU_KERNEL_T_S(SparseToDense,
+                      KernelAttr()
+                        .AddInputAttr(kNumberTypeInt64)
+                        .AddInputAttr(kNumberTypeUInt8)
+                        .AddInputAttr(kNumberTypeInt64)
+                        .AddOutputAttr(kNumberTypeUInt8),
+                      SparseToDenseCPUKernel, int64_t, uint8_t);
+
+MS_REG_CPU_KERNEL_T_S(SparseToDense,
+                      KernelAttr()
+                        .AddInputAttr(kNumberTypeInt64)
+                        .AddInputAttr(kNumberTypeInt16)
+                        .AddInputAttr(kNumberTypeInt64)
+                        .AddOutputAttr(kNumberTypeInt16),
+                      SparseToDenseCPUKernel, int64_t, int16_t);
+
+MS_REG_CPU_KERNEL_T_S(SparseToDense,
+                      KernelAttr()
+                        .AddInputAttr(kNumberTypeInt64)
+                        .AddInputAttr(kNumberTypeUInt16)
+                        .AddInputAttr(kNumberTypeInt64)
+                        .AddOutputAttr(kNumberTypeUInt16),
+                      SparseToDenseCPUKernel, int64_t, uint16_t);
+
+MS_REG_CPU_KERNEL_T_S(SparseToDense,
+                      KernelAttr()
+                        .AddInputAttr(kNumberTypeInt64)
+                        .AddInputAttr(kNumberTypeFloat16)
+                        .AddInputAttr(kNumberTypeInt64)
+                        .AddOutputAttr(kNumberTypeFloat16),
+                      SparseToDenseCPUKernel, int64_t, float16);
+
+MS_REG_CPU_KERNEL_T_S(SparseToDense,
+                      KernelAttr()
+                        .AddInputAttr(kNumberTypeInt64)
+                        .AddInputAttr(kNumberTypeFloat)
+                        .AddInputAttr(kNumberTypeInt64)
+                        .AddOutputAttr(kNumberTypeFloat),
+                      SparseToDenseCPUKernel, int64_t, float);
+
+MS_REG_CPU_KERNEL_T_S(SparseToDense,
+                      KernelAttr()
+                        .AddInputAttr(kNumberTypeInt64)
+                        .AddInputAttr(kNumberTypeFloat64)
+                        .AddInputAttr(kNumberTypeInt64)
+                        .AddOutputAttr(kNumberTypeFloat64),
+                      SparseToDenseCPUKernel, int64_t, double);
+
+MS_REG_CPU_KERNEL_T_S(SparseToDense,
+                      KernelAttr()
+                        .AddInputAttr(kNumberTypeInt64)
+                        .AddInputAttr(kNumberTypeBool)
+                        .AddInputAttr(kNumberTypeInt64)
+                        .AddOutputAttr(kNumberTypeBool),
+                      SparseToDenseCPUKernel, int64_t, bool);
+}  // namespace kernel
+}  // namespace mindspore
+#endif  // MINDSPORE_CCSRC_BACKEND_KERNEL_COMPILER_CPU_SPARSETODENSE_CPU_KERNEL_H_
diff --git a/tests/st/ops/cpu/test_sparse_to_dense_op.py b/tests/st/ops/cpu/test_sparse_to_dense_op.py
new file mode 100644
index 00000000000..702eb7928c5
--- /dev/null
+++ b/tests/st/ops/cpu/test_sparse_to_dense_op.py
@@ -0,0 +1,198 @@
+# Copyright 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.
+# You may obtain a copy of the License at
+#
+# http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+# ============================================================================
+
+import numpy as np
+import pytest
+import mindspore.context as context
+from mindspore import Tensor
+from mindspore.nn import Cell
+from mindspore.ops import operations as P
+
+context.set_context(mode=context.GRAPH_MODE, enable_graph_kernel=True, device_target="CPU")
+
+
+class SparseToDenseNet(Cell):
+    def __init__(self):
+        super(SparseToDenseNet, self).__init__()
+        self.sparse_to_dense = P.SparseToDense()
+
+    def construct(self, indices, values, dense_shape):
+        return self.sparse_to_dense(indices, values, dense_shape)
+
+
+@pytest.mark.level0
+@pytest.mark.platform_x86_cpu_training
+@pytest.mark.env_onecard
+def test_sparse_to_dense_A():
+    np.random.seed(0)
+    indices = np.array([[0, 1], [1, 2]]).astype(np.int32)
+    values = np.array([7, 8]).astype(np.int32)
+    dense_shape = (3, 4)
+    net = SparseToDenseNet()
+    result = net(Tensor(indices), Tensor(values), dense_shape)
+    expect_output = np.array([[0, 7, 0, 0],
+                              [0, 0, 8, 0],
+                              [0, 0, 0, 0]]).astype(np.int32)
+    assert np.allclose(result.asnumpy(), expect_output, rtol=1.e-4, atol=1.e-8, equal_nan=True)
+
+
+@pytest.mark.level0
+@pytest.mark.platform_x86_cpu_training
+@pytest.mark.env_onecard
+def test_sparse_to_dense_B():
+    np.random.seed(0)
+    indices = np.array([[0, 1], [1, 2], [2, 3]]).astype(np.int32)
+    values = np.array([6.5, 7.5, 9.5]).astype(np.float64)
+    dense_shape = (3, 4)
+    net = SparseToDenseNet()
+    result = net(Tensor(indices), Tensor(values), dense_shape)
+    expect_output = np.array([[0, 6.5, 0, 0],
+                              [0, 0, 7.5, 0],
+                              [0, 0, 0, 9.5]]).astype(np.float64)
+    assert np.allclose(result.asnumpy(), expect_output, rtol=1.e-4, atol=1.e-8, equal_nan=True)
+
+
+@pytest.mark.level0
+@pytest.mark.platform_x86_cpu_training
+@pytest.mark.env_onecard
+def test_sparse_to_dense_C():
+    np.random.seed(0)
+    indices = np.array([[0, 1, 0, 0],
+                        [1, 0, 0, 2],
+                        [2, 0, 3, 0],
+                        [4, 2, 3, 5]]).astype(np.int32)
+    values = np.array([26.5, 17.5, 39.5, 11.5]).astype(np.float16)
+    dense_shape = (10, 8, 5, 10)
+    net = SparseToDenseNet()
+    result = net(Tensor(indices), Tensor(values), dense_shape)
+    expect_output = np.zeros([10, 8, 5, 10]).astype(np.float16)
+    for i in range(0, indices.shape[0]):
+        j = indices[i][0]
+        k = indices[i][1]
+        l = indices[i][2]
+        m = indices[i][3]
+        expect_output[j][k][l][m] = values[i]
+    assert np.allclose(result.asnumpy(), expect_output, rtol=1.e-4, atol=1.e-8, equal_nan=True)
+
+
+@pytest.mark.level0
+@pytest.mark.platform_x86_cpu_training
+@pytest.mark.env_onecard
+def test_sparse_to_dense_D():
+    np.random.seed(0)
+    indices = np.array([[0, 1, 0, 0, 2, 1],
+                        [9, 0, 0, 8, 0, 0],
+                        [2, 0, 4, 0, 1, 1],
+                        [4, 2, 3, 5, 0, 2],
+                        [7, 4, 3, 9, 0, 1]]).astype(np.int32)
+    values = np.array([1, 1, 1, 1, 1]).astype(np.bool)
+    dense_shape = (10, 5, 5, 10, 3, 3)
+    net = SparseToDenseNet()
+    result = net(Tensor(indices), Tensor(values), dense_shape)
+    expect_output = np.zeros([10, 5, 5, 10, 3, 3]).astype(np.bool)
+    for i in range(0, indices.shape[0]):
+        j = indices[i][0]
+        k = indices[i][1]
+        l = indices[i][2]
+        m = indices[i][3]
+        u = indices[i][4]
+        v = indices[i][5]
+        expect_output[j][k][l][m][u][v] = values[i]
+    assert np.allclose(result.asnumpy(), expect_output, rtol=1.e-4, atol=1.e-8, equal_nan=True)
+
+
+@pytest.mark.level0
+@pytest.mark.platform_x86_cpu_training
+@pytest.mark.env_onecard
+def test_sparse_to_dense_E():
+    indices = np.array([2, 5, 7]).astype(np.int32)
+    values = np.array([17, 18, 19]).astype(np.int8)
+    dense_shape = ([10])
+    expect_output = np.zeros([10]).astype(np.int8)
+    for i in range(0, indices.shape[0]):
+        j = indices[i]
+        expect_output[j] = values[i]
+
+    net = SparseToDenseNet()
+    result = net(Tensor(indices), Tensor(values), dense_shape)
+    assert np.allclose(result.asnumpy(), expect_output, rtol=1.e-4, atol=1.e-8, equal_nan=True)
+
+
+@pytest.mark.level0
+@pytest.mark.platform_x86_cpu_training
+@pytest.mark.env_onecard
+def test_sparse_to_dense_F():
+    indices = np.array([2, 4, 18]).astype(np.int32)
+    values = np.array([-23, 18, -1]).astype(np.int16)
+    dense_shape = ([20])
+    expect_output = np.zeros([20]).astype(np.int16)
+    for i in range(0, indices.shape[0]):
+        j = indices[i]
+        expect_output[j] = values[i]
+
+    net = SparseToDenseNet()
+    result = net(Tensor(indices), Tensor(values), dense_shape)
+    assert np.allclose(result.asnumpy(), expect_output, rtol=1.e-4, atol=1.e-8, equal_nan=True)
+
+
+@pytest.mark.level0
+@pytest.mark.platform_x86_cpu_training
+@pytest.mark.env_onecard
+def test_sparse_to_dense_G():
+    indices = np.array([2, 5, 7]).astype(np.int32)
+    values = np.array([17, 18, 19]).astype(np.uint8)
+    dense_shape = ([10])
+    expect_output = np.zeros([10]).astype(np.uint8)
+    for i in range(0, indices.shape[0]):
+        j = indices[i]
+        expect_output[j] = values[i]
+
+    net = SparseToDenseNet()
+    result = net(Tensor(indices), Tensor(values), dense_shape)
+    assert np.allclose(result.asnumpy(), expect_output, rtol=1.e-4, atol=1.e-8, equal_nan=True)
+
+
+@pytest.mark.level0
+@pytest.mark.platform_x86_cpu_training
+@pytest.mark.env_onecard
+def test_sparse_to_dense_H():
+    indices = np.array([2, 5, 7]).astype(np.int32)
+    values = np.array([17, 18, 19]).astype(np.uint16)
+    dense_shape = ([10])
+    expect_output = np.zeros([10]).astype(np.uint16)
+    for i in range(0, indices.shape[0]):
+        j = indices[i]
+        expect_output[j] = values[i]
+
+    net = SparseToDenseNet()
+    result = net(Tensor(indices), Tensor(values), dense_shape)
+    assert np.allclose(result.asnumpy(), expect_output, rtol=1.e-4, atol=1.e-8, equal_nan=True)
+
+
+@pytest.mark.level0
+@pytest.mark.platform_x86_cpu_training
+@pytest.mark.env_onecard
+def test_sparse_to_dense_I():
+    indices = np.array([2, 5, 7]).astype(np.int64)
+    values = np.array([17, 18, 19]).astype(np.float16)
+    dense_shape = ([10])
+    expect_output = np.zeros([10]).astype(np.float16)
+    for i in range(0, indices.shape[0]):
+        j = indices[i]
+        expect_output[j] = values[i]
+
+    net = SparseToDenseNet()
+    result = net(Tensor(indices), Tensor(values), dense_shape)
+    assert np.allclose(result.asnumpy(), expect_output, rtol=1.e-4, atol=1.e-8, equal_nan=True)