!9559 add DynamicAssign, PadAndShift cpu ops

From: @fangzehua
Reviewed-by: 
Signed-off-by:

mindspore/ccsrc/backend/kernel_compiler/aicpu/aicpu_util.h

@@ -47,9 +47,12 @@ constexpr auto kEditDistance = "EditDistance";
 constexpr auto kGatherD = "GatherD";
 constexpr auto kIdentity = "Identity";
 constexpr auto kUpdateCache = "UpdateCache";
+constexpr auto kCacheSwapTable = "CacheSwapTable";
+constexpr auto kSubAndFilter = "SubAndFilter";
+constexpr auto kPadAndShift = "PadAndShift";
 constexpr auto kCustRunApi = "RunCpuKernel";
 const std::set<std::string> kCustAiCpuKernelOps{kEditDistance, kIdentity};
-const std::set<std::string> kCacheKernelOps{kUpdateCache};
+const std::set<std::string> kCacheKernelOps{kUpdateCache, kCacheSwapTable, kSubAndFilter, kPadAndShift};
 
 struct AicpuParamHead {
   uint32_t length;         // Total length: include cunstom message

mindspore/ccsrc/backend/kernel_compiler/cpu/dynamic_assign_cpu_kernel.cc

@@ -0,0 +1,89 @@
+/**
+ * Copyright 2020 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 <algorithm>
+#include "backend/kernel_compiler/cpu/dynamic_assign_cpu_kernel.h"
+#include "runtime/device/cpu/cpu_device_address.h"
+
+namespace mindspore {
+namespace kernel {
+void DynamicAssignCPUKernel::InitKernel(const CNodePtr &kernel_node) {
+  node_ = kernel_node;
+  input_x_dtype_ = AnfAlgo::GetPrevNodeOutputInferDataType(kernel_node, 0);
+  input_x_dtype_size_ = GetTypeByte(TypeIdToType(input_x_dtype_));
+}
+
+bool DynamicAssignCPUKernel::Launch(const std::vector<kernel::AddressPtr> &inputs,
+                                    const std::vector<kernel::AddressPtr> & /*workspace*/,
+                                    const std::vector<kernel::AddressPtr> &outputs) {
+  if (input_x_dtype_ == kNumberTypeInt32) {
+    LaunchKernel<int>(inputs, outputs);
+  } else if (input_x_dtype_ == kNumberTypeInt64) {
+    LaunchKernel<int64_t>(inputs, outputs);
+  } else if (input_x_dtype_ == kNumberTypeFloat32) {
+    LaunchKernel<float>(inputs, outputs);
+  } else if (input_x_dtype_ == kNumberTypeFloat64) {
+    LaunchKernel<double>(inputs, outputs);
+  } else {
+    MS_LOG(ERROR) << "Dtype of indices only support float32, float64, int32, int64";
+    return false;
+  }
+  return true;
+}
+
+template <typename T>
+void DynamicAssignCPUKernel::LaunchKernel(const std::vector<AddressPtr> &inputs,
+                                          const std::vector<kernel::AddressPtr> &outputs) {
+  auto input_x_shape = AnfAlgo::GetPrevNodeOutputInferShape(node_, 0);
+  auto input_y_shape = AnfAlgo::GetPrevNodeOutputInferShape(node_, 1);
+  batch_size_ = 1;
+  for (size_t i = 0; i < input_x_shape.size(); ++i) {
+    batch_size_ *= input_x_shape[i];
+  }
+
+  if (input_x_shape.size() != input_y_shape.size()) MS_LOG(EXCEPTION) << "x y must be same shape";
+  for (size_t i = 0; i < input_x_shape.size(); ++i) {
+    if (input_x_shape[i] != input_y_shape[i]) {
+      MS_LOG(EXCEPTION) << "x y must be same shape";
+    }
+  }
+  T *input_x = reinterpret_cast<T *>(inputs[0]->addr);
+  T *input_y = reinterpret_cast<T *>(inputs[1]->addr);
+  auto max_size = inputs[0]->size;
+  size_t total_size = input_x_dtype_size_ * batch_size_;
+  if (total_size > max_size) {
+    MS_LOG(EXCEPTION) << "Memcpy size must <= max_size, but got memcpy size is : " << total_size
+                      << ", max size is : " << max_size;
+  }
+  int ret = memcpy_s(input_x, total_size, input_y, total_size);
+  if (ret != 0) {
+    MS_LOG(EXCEPTION) << "Memcpy_s error, errorno" << ret;
+  }
+
+  auto node_with_idx = AnfAlgo::GetPrevNodeOutput(node_, 0);
+  auto node = node_with_idx.first;
+  if (node->isa<Parameter>()) {
+    auto node_ptr = node->cast<ParameterPtr>();
+    auto value = node_ptr->default_param();
+    auto tensor = value->cast<std::shared_ptr<tensor::Tensor>>();
+    ShapeVector shape_tmp;
+    (void)std::transform(input_x_shape.begin(), input_x_shape.end(), std::back_inserter(shape_tmp), SizeToLong);
+    tensor->set_shape(shape_tmp);
+  } else {
+    MS_LOG(EXCEPTION) << "Input x must be a Parameter.";
+  }
+}
+}  // namespace kernel
+}  // namespace mindspore

mindspore/ccsrc/backend/kernel_compiler/cpu/dynamic_assign_cpu_kernel.h

@@ -0,0 +1,69 @@
+/**
+ * Copyright 2020 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_DYNAMIC_ASSIGN_CPU_KERNEL_H_
+#define MINDSPORE_CCSRC_BACKEND_KERNEL_COMPILER_CPU_DYNAMIC_ASSIGN_CPU_KERNEL_H_
+
+#include <vector>
+#include <memory>
+#include <unordered_map>
+#include "backend/kernel_compiler/cpu/cpu_kernel.h"
+#include "backend/kernel_compiler/cpu/cpu_kernel_factory.h"
+
+namespace mindspore {
+namespace kernel {
+class DynamicAssignCPUKernel : public CPUKernel {
+ public:
+  DynamicAssignCPUKernel() = default;
+  ~DynamicAssignCPUKernel() 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;
+
+  template <typename T>
+  void LaunchKernel(const std::vector<AddressPtr> &inputs, const std::vector<kernel::AddressPtr> &outputs);
+
+ private:
+  size_t batch_size_{1};
+  TypeId input_x_dtype_{kTypeUnknown};
+  size_t input_x_dtype_size_ = 4;
+  CNodePtr node_ = nullptr;
+};
+
+MS_REG_CPU_KERNEL(
+  DynamicAssign,
+  KernelAttr().AddInputAttr(kNumberTypeInt32).AddInputAttr(kNumberTypeInt32).AddOutputAttr(kNumberTypeInt32),
+  DynamicAssignCPUKernel);
+
+MS_REG_CPU_KERNEL(
+  DynamicAssign,
+  KernelAttr().AddInputAttr(kNumberTypeInt64).AddInputAttr(kNumberTypeInt64).AddOutputAttr(kNumberTypeInt64),
+  DynamicAssignCPUKernel);
+
+MS_REG_CPU_KERNEL(
+  DynamicAssign,
+  KernelAttr().AddInputAttr(kNumberTypeFloat32).AddInputAttr(kNumberTypeFloat32).AddOutputAttr(kNumberTypeFloat32),
+  DynamicAssignCPUKernel);
+
+MS_REG_CPU_KERNEL(
+  DynamicAssign,
+  KernelAttr().AddInputAttr(kNumberTypeFloat64).AddInputAttr(kNumberTypeFloat64).AddOutputAttr(kNumberTypeFloat64),
+  DynamicAssignCPUKernel);
+}  // namespace kernel
+}  // namespace mindspore
+
+#endif  // MINDSPORE_CCSRC_BACKEND_KERNEL_COMPILER_CPU_DYNAMIC_ASSIGN_CPU_KERNEL_H_

mindspore/ccsrc/backend/kernel_compiler/cpu/pad_and_shift_cpu_kernel.cc

@@ -0,0 +1,87 @@
+/**
+ * Copyright 2020 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/pad_and_shift_cpu_kernel.h"
+#include <string>
+#include "runtime/device/cpu/cpu_device_address.h"
+
+namespace mindspore {
+namespace kernel {
+void PadAndShiftCPUKernel::InitKernel(const CNodePtr &kernel_node) {
+  MS_EXCEPTION_IF_NULL(kernel_node);
+  node_ = kernel_node;
+  input_x_dtype_ = AnfAlgo::GetPrevNodeOutputInferDataType(kernel_node, 0);
+  type_size_ = GetTypeByte(TypeIdToType(input_x_dtype_));
+  auto indices_shape = AnfAlgo::GetPrevNodeOutputInferShape(kernel_node, 0);
+  batch_size_ = 1;
+  for (size_t i = 0; i < indices_shape.size(); ++i) {
+    batch_size_ *= indices_shape[i];
+  }
+  MS_LOG(INFO) << "PadAndShift batch_size:" << batch_size_;
+  auto cum_sum_arr_shape = AnfAlgo::GetPrevNodeOutputInferShape(kernel_node, 1);
+  if (cum_sum_arr_shape.size() != 1) {
+    MS_LOG(ERROR) << "The shape of cum_sum_arr must be 1.";
+  }
+  cum_sum_size_ = cum_sum_arr_shape[0];
+}
+
+bool PadAndShiftCPUKernel::Launch(const std::vector<kernel::AddressPtr> &inputs,
+                                  const std::vector<kernel::AddressPtr> & /*workspace*/,
+                                  const std::vector<kernel::AddressPtr> &outputs) {
+  if (input_x_dtype_ == kNumberTypeInt32) {
+    LaunchKernel<int>(inputs, outputs);
+  } else if (input_x_dtype_ == kNumberTypeInt64) {
+    LaunchKernel<int64_t>(inputs, outputs);
+  } else {
+    MS_LOG(ERROR) << "Dtype of input_x only support int32, int64";
+    return false;
+  }
+  return true;
+}
+
+template <typename T>
+void PadAndShiftCPUKernel::LaunchKernel(const std::vector<AddressPtr> &inputs,
+                                        const std::vector<kernel::AddressPtr> &outputs) {
+  T *input_x = reinterpret_cast<T *>(inputs[0]->addr);
+  T *cum_sum_arr = reinterpret_cast<T *>(inputs[1]->addr);
+  T shift_idx = *reinterpret_cast<T *>(inputs[2]->addr);
+  T *output = reinterpret_cast<T *>(outputs[0]->addr);
+
+  if (shift_idx >= static_cast<T>(cum_sum_size_)) {
+    MS_LOG(EXCEPTION) << "Shift index must small than cumsum size.";
+  }
+  size_t output_size = cum_sum_arr[cum_sum_size_ - 1];
+  T shift_size = cum_sum_arr[shift_idx];
+  T valid_size = cum_sum_arr[shift_idx + 1] - shift_size;
+  int ret = memset_s(output, outputs[0]->size, -1, type_size_ * output_size);
+  if (ret != 0) {
+    MS_LOG(EXCEPTION) << "memset_s error, errorno" << ret;
+  }
+  ret = memcpy_s(output + shift_size, valid_size * type_size_, input_x, valid_size * type_size_);
+  if (ret != 0) {
+    MS_LOG(EXCEPTION) << "memcpy_s error, errorno" << ret;
+  }
+  std::vector<size_t> out_shape;
+  out_shape.emplace_back(output_size);
+  std::vector<TypeId> dtypes;
+  auto output_nums = AnfAlgo::GetOutputTensorNum(node_);
+  for (size_t i = 0; i < output_nums; i++) {
+    dtypes.push_back(AnfAlgo::GetOutputInferDataType(node_, i));
+  }
+  AnfAlgo::SetOutputInferTypeAndShape(dtypes, {out_shape}, node_.get());
+}
+}  // namespace kernel
+}  // namespace mindspore

mindspore/ccsrc/backend/kernel_compiler/cpu/pad_and_shift_cpu_kernel.h

@@ -0,0 +1,66 @@
+/**
+ * Copyright 2020 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_PAD_AND_SHIFT_CPU_KERNEL_H_
+#define MINDSPORE_CCSRC_BACKEND_KERNEL_COMPILER_CPU_PAD_AND_SHIFT_CPU_KERNEL_H_
+
+#include <vector>
+#include <memory>
+#include <unordered_map>
+#include "backend/kernel_compiler/cpu/cpu_kernel.h"
+#include "backend/kernel_compiler/cpu/cpu_kernel_factory.h"
+
+namespace mindspore {
+namespace kernel {
+class PadAndShiftCPUKernel : public CPUKernel {
+ public:
+  PadAndShiftCPUKernel() = default;
+  ~PadAndShiftCPUKernel() 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;
+
+  template <typename T>
+  void LaunchKernel(const std::vector<AddressPtr> &inputs, const std::vector<kernel::AddressPtr> &outputs);
+
+ private:
+  size_t batch_size_{1};
+  size_t cum_sum_size_{1};
+  size_t type_size_{4};
+  TypeId input_x_dtype_{kTypeUnknown};
+  CNodePtr node_ = nullptr;
+};
+
+MS_REG_CPU_KERNEL(PadAndShift,
+                  KernelAttr()
+                    .AddInputAttr(kNumberTypeInt32)
+                    .AddInputAttr(kNumberTypeInt32)
+                    .AddInputAttr(kNumberTypeInt32)
+                    .AddOutputAttr(kNumberTypeInt32),
+                  PadAndShiftCPUKernel);
+
+MS_REG_CPU_KERNEL(PadAndShift,
+                  KernelAttr()
+                    .AddInputAttr(kNumberTypeInt64)
+                    .AddInputAttr(kNumberTypeInt64)
+                    .AddInputAttr(kNumberTypeInt64)
+                    .AddOutputAttr(kNumberTypeInt64),
+                  PadAndShiftCPUKernel);
+}  // namespace kernel
+}  // namespace mindspore
+
+#endif  // MINDSPORE_CCSRC_BACKEND_KERNEL_COMPILER_CPU_PAD_AND_SHIFT_CPU_KERNEL_H_

mindspore/ccsrc/utils/utils.h

@@ -263,6 +263,8 @@ constexpr auto kDropoutOpName = "Dropout";
 constexpr auto kDropoutGradOpName = "DropoutGrad";
 constexpr auto kDropoutGenMaskOpName = "DropoutGenMask";
 constexpr auto kDropoutDoMaskOpName = "DropoutDoMask";
+constexpr auto kSubAndFilterOpName = "SubAndFilter";
+constexpr auto kPadAndShiftOpName = "PadAndShift";
 constexpr auto kSparseSoftmaxCrossEntropyWithLogitsOpName = "SparseSoftmaxCrossEntropyWithLogits";
 constexpr auto kOneHotOpName = "OneHot";
 constexpr auto kSoftmaxCrossEntropyWithLogitsOpName = "SoftmaxCrossEntropyWithLogits";
@@ -482,7 +484,8 @@ const std::set<std::string> kHWSpecialFormatSet = {
 
 const std::set<TypeId> kFloatDataTypeSet = {kNumberTypeFloat16, kNumberTypeFloat32};
 
-const std::set<std::string> kComputeDepend = {kUniqueOpName, kComputeAccidentalHitsOpName};
+const std::set<std::string> kComputeDepend = {kUniqueOpName, kComputeAccidentalHitsOpName, kSubAndFilterOpName,
+                                              kPadAndShiftOpName};
 
 static inline void ChangeFileMode(const std::string &file_name, mode_t mode) {
   try {

mindspore/core/abstract/infer_functions.h

@@ -223,6 +223,10 @@ AbstractBasePtr InferImplUpdateCache(const AnalysisEnginePtr &, const PrimitiveP
                                      const AbstractBasePtrList &args_spec_list);
 AbstractBasePtr InferImplComputeAccidentalHits(const AnalysisEnginePtr &, const PrimitivePtr &primitive,
                                                const AbstractBasePtrList &args_spec_list);
+AbstractBasePtr InferImplPadAndShift(const AnalysisEnginePtr &, const PrimitivePtr &primitive,
+                                     const AbstractBasePtrList &args_spec_list);
+AbstractBasePtr InferImplDynamicAssign(const AnalysisEnginePtr &, const PrimitivePtr &primitive,
+                                       const AbstractBasePtrList &args_spec_list);
 AbstractBasePtr InferImplGatherV2(const AnalysisEnginePtr &, const PrimitivePtr &primitive,
                                   const AbstractBasePtrList &args_spec_list);
 AbstractBasePtr InferImplShape(const AnalysisEnginePtr &, const PrimitivePtr &primitive,

mindspore/core/abstract/prim_arrays.cc

@@ -193,6 +193,27 @@ AbstractBasePtr InferImplUnique(const AnalysisEnginePtr &, const PrimitivePtr &p
   return std::make_shared<AbstractTuple>(elements);
 }
 
+AbstractBasePtr InferImplPadAndShift(const AnalysisEnginePtr &, const PrimitivePtr &primitive,
+                                     const AbstractBasePtrList &args_spec_list) {
+  // inputs: a 1-d Tensor
+  const std::string op_name = primitive->name();
+  CheckArgsSize(op_name, args_spec_list, 3);
+  AbstractTensorPtr input = CheckArg<AbstractTensor>(op_name, args_spec_list, 0);
+  MS_EXCEPTION_IF_NULL(input);
+  auto shape = input->shape();
+  MS_EXCEPTION_IF_NULL(shape);
+  if (shape->shape().size() != 1) {
+    MS_LOG(EXCEPTION) << "Rank of " << op_name << "'s input must be 1.";
+  }
+  ShapeVector ids_shape = {Shape::SHP_ANY};
+  ShapeVector min_shape = {1};
+  ShapeVector max_shape = shape->max_shape();
+  if (max_shape.empty()) {
+    max_shape = shape->shape();
+  }
+  return std::make_shared<AbstractTensor>(input->element(), std::make_shared<Shape>(ids_shape, min_shape, max_shape));
+}
+
 AbstractBasePtr InferImplUniqueGrad(const AnalysisEnginePtr &, const PrimitivePtr &primitive,
                                     const AbstractBasePtrList &args_spec_list) {
   // inputs: a 1-d Tensor
@@ -612,6 +633,29 @@ AbstractBasePtr InferImplGatherV2(const AnalysisEnginePtr &, const PrimitivePtr
   return std::make_shared<AbstractTensor>(params->element(), std::make_shared<Shape>(out_shape));
 }
 
+AbstractBasePtr InferImplDynamicAssign(const AnalysisEnginePtr &, const PrimitivePtr &primitive,
+                                       const AbstractBasePtrList &args_spec_list) {
+  // Inputs: a tensor
+  CheckArgsSize(primitive->name(), args_spec_list, 2);
+
+  MS_LOG(INFO) << "InferImplDynamicAssign " << args_spec_list[0];
+  auto type = args_spec_list[0]->BuildType();
+  if (type->type_id() == kObjectTypeRefKey) {
+    return args_spec_list[1]->Broaden();
+  } else {
+    auto x = CheckArg<AbstractTensor>(primitive->name(), args_spec_list, 0);
+    auto y = CheckArg<AbstractTensor>(primitive->name(), args_spec_list, 1);
+    MS_EXCEPTION_IF_NULL(x);
+    MS_EXCEPTION_IF_NULL(y);
+    auto y_shape = y->shape();
+    MS_EXCEPTION_IF_NULL(y_shape);
+    if (!y_shape->max_shape().empty()) {
+      x->set_shape(y->shape());
+    }
+    return args_spec_list[0];
+  }
+}
+
 AbstractBasePtr InferImplEmbeddingLookup(const AnalysisEnginePtr &, const PrimitivePtr &primitive,
                                          const AbstractBasePtrList &args_spec_list) {
   const std::string op_name = primitive->name();

mindspore/core/abstract/primitive_infer_map.cc

@@ -67,9 +67,11 @@ PrimitiveEvalImplMap &GetPrimitiveToEvalImplMap() {
     {prim::kPrimSubAndFilter, {InferImplSubAndFilter, true}},
     {prim::kPrimScatterUpdate, {InferImplScatterUpdate, true}},
     {prim::kPrimMapCacheIdx, {InferImplMapCacheIdx, true}},
+    {prim::kPrimDynamicAssign, {InferImplDynamicAssign, true}},
     {prim::kPrimCacheSwapTable, {InferImplCacheSwapTable, true}},
     {prim::kPrimUpdateCache, {InferImplUpdateCache, true}},
     {prim::kPrimComputeAccidentalHits, {InferImplComputeAccidentalHits, true}},
+    {prim::kPrimPadAndShift, {InferImplPadAndShift, true}},
     {prim::kPrimDiv, {InferImplDiv, true}},
     {prim::kPrimRealDiv, {InferImplRealDiv, true}},
     {prim::kPrimShape, {InferImplShape, false}},

mindspore/core/base/core_ops.h

@@ -103,6 +103,8 @@ inline const PrimitivePtr kPrimMapCacheIdx = std::make_shared<Primitive>("MapCac
 inline const PrimitivePtr kPrimUpdateCache = std::make_shared<Primitive>("UpdateCache");
 inline const PrimitivePtr kPrimComputeAccidentalHits = std::make_shared<Primitive>("ComputeAccidentalHits");
 inline const PrimitivePtr kPrimCacheSwapTable = std::make_shared<Primitive>("CacheSwapTable");
+inline const PrimitivePtr kPrimDynamicAssign = std::make_shared<Primitive>("DynamicAssign");
+inline const PrimitivePtr kPrimPadAndShift = std::make_shared<Primitive>("PadAndShift");
 inline const PrimitivePtr kPrimSlice = std::make_shared<Primitive>("Slice");
 inline const PrimitivePtr kPrimTile = std::make_shared<Primitive>("Tile");
 inline const PrimitivePtr kPrimAddN = std::make_shared<Primitive>("AddN");

mindspore/ops/_op_impl/aicpu/__init__.py

@@ -24,6 +24,8 @@ from .scatter import _scatter_aicpu
 from .identity import _identity_aicpu
 from .edit_distance import _edit_distance_aicpu
 from .unique_with_pad import _unique_with_pad_aicpu
+from .sub_and_filter import _sub_and_filter_aicpu
+from .pad_and_shift import _pad_and_shift_aicpu
 from .dropout_genmask import _dropout_genmask_aicpu
 from .get_next import _get_next_aicpu
 from .print_tensor import _print_aicpu

mindspore/ops/_op_impl/aicpu/pad_and_shift.py

@@ -0,0 +1,33 @@
+# Copyright 2020 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.
+# ============================================================================
+
+"""PadAndShift op"""
+from mindspore.ops.op_info_register import op_info_register, AiCPURegOp, DataType
+
+pad_and_shift_op_info = AiCPURegOp("PadAndShift") \
+    .fusion_type("OPAQUE") \
+    .input(0, "x", "required") \
+    .input(1, "cum_sum_arr", "required") \
+    .input(2, "shift_idx", "required") \
+    .output(0, "output", "required") \
+    .dtype_format(DataType.I32_Default, DataType.I32_Default, DataType.I32_Default, DataType.I32_Default) \
+    .dtype_format(DataType.I64_Default, DataType.I64_Default, DataType.I64_Default, DataType.I64_Default) \
+    .get_op_info()
+
+
+@op_info_register(pad_and_shift_op_info)
+def _pad_and_shift_aicpu():
+    """PadAndShift AiCPU register"""
+    return

mindspore/ops/_op_impl/aicpu/sub_and_filter.py

@@ -0,0 +1,36 @@
+# Copyright 2020 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.
+# ============================================================================
+
+"""SubAndFilter op"""
+from mindspore.ops.op_info_register import op_info_register, AiCPURegOp, DataType
+
+sub_and_filter_op_info = AiCPURegOp("SubAndFilter") \
+    .fusion_type("OPAQUE") \
+    .input(0, "x", "required") \
+    .input(1, "max_num", "required") \
+    .input(2, "offset", "required") \
+    .output(0, "filter_res", "required") \
+    .output(1, "filter_idx", "required") \
+    .dtype_format(DataType.I32_Default, DataType.I32_Default,
+                  DataType.I32_Default, DataType.I32_Default, DataType.I32_Default) \
+    .dtype_format(DataType.I64_Default, DataType.I64_Default,
+                  DataType.I64_Default, DataType.I64_Default, DataType.I64_Default) \
+    .get_op_info()
+
+
+@op_info_register(sub_and_filter_op_info)
+def _sub_and_filter_aicpu():
+    """SubAndFilter AiCPU register"""
+    return

mindspore/ops/operations/__init__.py

@@ -92,8 +92,8 @@ from ._thor_ops import (CusBatchMatMul, CusCholeskyTrsm, CusFusedAbsMax1, CusImg
                         CusMatMulCubeFraczLeftCast, Im2Col, UpdateThorGradient, Cholesky, CholeskyTrsm, DetTriangle,
                         ProdForceSeA)
 from .sparse_ops import SparseToDense
-from ._cache_ops import (CacheSwapHashmap, SearchCacheIdx, CacheSwapTable, UpdateCache, MapCacheIdx, SubAndFilter,
+from ._embedding_cache_ops import (CacheSwapHashmap, SearchCacheIdx, CacheSwapTable, UpdateCache, MapCacheIdx, SubAndFilter,
-                         MapUniform)
+                                   MapUniform, DynamicAssign, PadAndShift)
 
 __all__ = [
     'Unique',

mindspore/ops/operations/_embedding_cache_ops.py

@@ -93,7 +93,7 @@ class SubAndFilter(PrimitiveWithCheck):
                                 outputs=['sub_res', 'sub_idx'])
 
     def check_shape(self, input_x_shape, max_num_shape, offset_shape):
-        return (-1, -1)
+        return ((-1,), (-1,))
 
     def check_dtype(self, input_x_dtype, max_num_dtype, offset_dtype):
         validator.check_tensor_dtype_valid(
@@ -358,3 +358,77 @@ class MapCacheIdx(PrimitiveWithCheck):
         else:
             out['min_shape'] = (0, 0, 0, 0)
         return out
+
+
+class DynamicAssign(PrimitiveWithCheck):
+    """
+    Assigns `Parameter` with a value, the `value` can have a dynamic shape.
+
+    Inputs:
+        - **variable** (Parameter) - The `Parameter`.
+        - **value** (Tensor) - The value to be assigned.
+
+    Outputs:
+        Tensor, has the same type as original `variable`.
+
+    Supported Platforms:
+        `CPU`
+    """
+    __mindspore_signature__ = (
+        sig.make_sig('variable', sig.sig_rw.RW_WRITE, dtype=sig.sig_dtype.T),
+        sig.make_sig('value', dtype=sig.sig_dtype.T)
+    )
+
+    @prim_attr_register
+    def __init__(self):
+        self.init_prim_io_names(inputs=['ref', 'value'], outputs=['output'])
+
+    def check_dtype(self, variable, value):
+        if variable != mstype.type_refkey:
+            validator.check_tensor_dtype_valid(
+                "variable", variable, mstype.number_type, self.name)
+        validator.check_scalar_or_tensor_types_same(
+            {"value": value}, mstype.number_type, self.name)
+
+
+class PadAndShift(PrimitiveWithCheck):
+    """
+    Pad a tensor with -1, and shift with a length.
+
+    Inputs:
+        - **input_x** (Tensor) - The input Tensor, which will be copyed
+            to `output`.
+        - **cum_sum_arr** (Tensor) - The last value of cum_sum_arr is
+            the pad length of output tensor, cum_sum_arr[shift_idx] is
+            the start to shift, and cum_sum_arr[shift_idx+1] is the end.
+        - **shift_idx** (Int) - The idx of cum_sum_arr.
+        if use python, PadAndShift is:
+            output = [-1] * cum_sum_arr[-1]
+            start = cum_sum_arr[shift_idx]
+            end = cum_sum_arr[shift_idx + 1]
+            output[start:end] = input_x[:(end-start)]
+    Outputs:
+        Tensor, has the same type as original `variable`.
+
+    Supported Platforms:
+        `CPU`
+
+    Examples:
+        >>> input_x = Tensor(np.array([9, 13, -1, -1, -1, -1, -1, -1]), mstype.int32)
+        >>> cum_sum_arr = Tensor(np.array([0, 3, 5]), mstype.int32)
+        >>> shift_idx = 1
+        >>> pad_and_shift = ops.PadAndShift()
+        >>> output = pad_and_shift(input_x, cum_sum_arr, shift_idx)
+        >>> print(output)
+        [-1, -1, -1, 9, 13]
+    """
+    @prim_attr_register
+    def __init__(self):
+        self.init_prim_io_names(
+            inputs=['input_x', 'cum_sum_arr', 'shift_idx'], outputs=['output'])
+
+    def check_shape(self, input_x_shape, cum_sum_arr_shape, shift_idx_shape):
+        return input_x_shape
+
+    def check_dtype(self, input_x_dtype, cum_sum_arr_dtype, shift_idx_dtype):
+        return input_x_dtype

tests/st/dynamic_shape/test_dynamic_assign_cpu.py

@@ -0,0 +1,49 @@
+# Copyright 2020 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
+import mindspore.nn as nn
+from mindspore import Tensor, Parameter
+import mindspore.common.dtype as mstype
+from mindspore.ops import operations as P
+
+context.set_context(mode=context.GRAPH_MODE, device_target="CPU")
+
+
+class Net(nn.Cell):
+    def __init__(self):
+        super(Net, self).__init__()
+        self.unique = P.Unique()
+        self.dynamic_assign = P.DynamicAssign()
+        self.param = Parameter(
+            Tensor(np.zeros((5,), np.int32)), name="assign_x")
+
+    def construct(self, y):
+        y, _ = self.unique(y)
+        return self.dynamic_assign(self.param, y)
+
+
+@pytest.mark.level0
+@pytest.mark.platform_arm_ascend_training
+@pytest.mark.platform_x86_ascend_training
+@pytest.mark.env_onecard
+def test_dynamic_assign():
+    y = Tensor(np.array([2, 2, 3, 3, 4]), mstype.int32)
+    dynamic_assign = Net()
+    _ = dynamic_assign(y)
+    expect1 = np.array([2, 3, 4])
+    param_np = dynamic_assign.param.data.asnumpy()
+    assert (param_np == expect1).all()

tests/st/dynamic_shape/test_pad_and_shift_cpu.py

@@ -0,0 +1,46 @@
+# Copyright 2020 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
+import mindspore.nn as nn
+from mindspore import Tensor
+import mindspore.common.dtype as mstype
+from mindspore.ops import operations as P
+
+context.set_context(mode=context.GRAPH_MODE, device_target="CPU")
+
+
+class Net(nn.Cell):
+    def __init__(self):
+        super(Net, self).__init__()
+        self.pad_and_shift = P.PadAndShift()
+        self.shift_idx = 1
+
+    def construct(self, x, y):
+        return self.pad_and_shift(x, y, self.shift_idx)
+
+
+@pytest.mark.level0
+@pytest.mark.platform_arm_ascend_training
+@pytest.mark.platform_x86_ascend_training
+@pytest.mark.env_onecard
+def test_pad_and_shift_cpu():
+    x = Tensor(np.array([9, 13, -1, -1, -1, -1, -1, -1]), mstype.int32)
+    y = Tensor(np.array([0, 3, 5]), mstype.int32)
+    net = Net()
+    output = net(x, y)
+    expect = np.array([-1, -1, -1, 9, 13])
+    assert (output.asnumpy() == expect).all()