• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 /**
2  * Copyright 2022 Huawei Technologies Co., Ltd
3  *
4  * Licensed under the Apache License, Version 2.0 (the "License");
5  * you may not use this file except in compliance with the License.
6  * You may obtain a copy of the License at
7  *
8  * http://www.apache.org/licenses/LICENSE-2.0
9  *
10  * Unless required by applicable law or agreed to in writing, software
11  * distributed under the License is distributed on an "AS IS" BASIS,
12  * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13  * See the License for the specific language governing permissions and
14  * limitations under the License.
15  */
16 
17 #include "ops/sparse_segment_sum.h"
18 
19 #include <map>
20 #include <memory>
21 #include <set>
22 
23 #include "abstract/ops/primitive_infer_map.h"
24 #include "mindapi/src/helper.h"
25 #include "mindspore/core/ops/sparse_ops.h"
26 #include "ops/op_name.h"
27 
28 namespace mindspore {
29 namespace ops {
30 namespace {
SparseSegmentSumInferShape(const PrimitivePtr & prim,const std::vector<AbstractBasePtr> & input_args)31 abstract::ShapePtr SparseSegmentSumInferShape(const PrimitivePtr &prim,
32                                               const std::vector<AbstractBasePtr> &input_args) {
33   MS_EXCEPTION_IF_NULL(prim);
34   auto prim_name = prim->name();
35   auto x_shape = CheckAndConvertUtils::ConvertShapePtrToShapeMap(input_args[kInputIndex0]->GetShape())[kShape];
36   auto indices_shape = CheckAndConvertUtils::ConvertShapePtrToShapeMap(input_args[kInputIndex1]->GetShape())[kShape];
37   auto segment_ids_shape =
38     CheckAndConvertUtils::ConvertShapePtrToShapeMap(input_args[kInputIndex2]->GetShape())[kShape];
39   // support dynamic rank
40   if (IsDynamicRank(x_shape) || IsDynamicRank(indices_shape) || IsDynamicRank(segment_ids_shape)) {
41     return std::make_shared<abstract::Shape>(ShapeVector({abstract::Shape::kShapeRankAny}));
42   }
43   (void)CheckAndConvertUtils::CheckInteger("indices_shape", SizeToLong(indices_shape.size()), kEqual,
44                                            SizeToLong(kInputIndex1), prim->name());
45   (void)CheckAndConvertUtils::CheckInteger("segment_ids_shape", SizeToLong(segment_ids_shape.size()), kEqual,
46                                            SizeToLong(kInputIndex1), prim->name());
47   if (x_shape.size() < kInputIndex1) {
48     MS_EXCEPTION(ValueError) << "For '" << prim_name << "', "
49                              << "x's rank must be greater than 1, but got [" << x_shape.size() << "].";
50   }
51   if (!(IsDynamic(indices_shape) || IsDynamic(segment_ids_shape)) &&
52       indices_shape[kInputIndex0] != segment_ids_shape[kInputIndex0]) {
53     MS_EXCEPTION(ValueError) << "For '" << prim_name << "', the rank of indices and segment_ids must be the same, "
54                              << "but got indices [" << indices_shape[kInputIndex0] << "] "
55                              << "and segment_ids [" << segment_ids_shape[kInputIndex0] << "].";
56   }
57   if ((indices_shape[kInputIndex0] == kInputIndex0) || (segment_ids_shape[kInputIndex0] == kInputIndex0)) {
58     MS_EXCEPTION(ValueError) << "For '" << prim_name << "', the rank of indices and segment_ids must greater than 0, "
59                              << "but got indices [" << indices_shape[kInputIndex0] << "] "
60                              << "and segment_ids [" << segment_ids_shape[kInputIndex0] << "].";
61   }
62   if (!input_args[kInputIndex2]->GetValue()->isa<ValueAny>() && !input_args[kInputIndex2]->GetValue()->isa<None>()) {
63     auto segment_ids_value_ptr = input_args[kInputIndex2]->GetValue();
64     MS_EXCEPTION_IF_NULL(segment_ids_value_ptr);
65     auto segment_ids_type_ptr = input_args[kInputIndex2]->GetType();
66     MS_EXCEPTION_IF_NULL(segment_ids_type_ptr);
67     auto segment_ids_value_ptr_tensor = CheckAndConvertUtils::CheckTensorIntValue("segment_ids", segment_ids_value_ptr,
68                                                                                   prim->name(), segment_ids_type_ptr);
69     size_t dim_zero = static_cast<size_t>(segment_ids_value_ptr_tensor.back()) + kInputIndex1;
70     if (dim_zero < kInputIndex1) {
71       MS_EXCEPTION(ValueError) << "For '" << prim_name << "', segment_ids must be greater or equal to 0, "
72                                << "but got [" << dim_zero << "].";
73     } else {
74       ShapeVector y_shape = x_shape;
75       y_shape[kInputIndex0] = static_cast<int64_t>(dim_zero);
76       return std::make_shared<abstract::Shape>(y_shape);
77     }
78   } else {
79     ShapeVector output_shape = x_shape;
80     output_shape[kInputIndex0] = -1;
81     return std::make_shared<abstract::Shape>(output_shape);
82   }
83 }
84 
SparseSegmentSumInferType(const PrimitivePtr & prim,const std::vector<AbstractBasePtr> & input_args)85 TypePtr SparseSegmentSumInferType(const PrimitivePtr &prim, const std::vector<AbstractBasePtr> &input_args) {
86   MS_EXCEPTION_IF_NULL(prim);
87   auto prim_name = prim->name();
88   auto x_type = input_args[kInputIndex0]->GetType();
89   auto indices_type = input_args[kInputIndex1]->GetType();
90   auto segment_ids_type = input_args[kInputIndex2]->GetType();
91   const std::set<TypePtr> valid_types = {kInt8, kInt16, kInt32, kInt64, kUInt8, kUInt16, kFloat16, kFloat32, kFloat64};
92   const std::set<TypePtr> common_valid_types = {kInt32, kInt64};
93   (void)CheckAndConvertUtils::CheckTensorTypeValid("x", x_type, valid_types, prim_name);
94   std::map<std::string, TypePtr> types;
95   (void)types.emplace("indices", indices_type);
96   (void)types.emplace("segment_ids", segment_ids_type);
97   (void)CheckAndConvertUtils::CheckTensorTypeSame(types, common_valid_types, prim->name());
98   return input_args[kInputIndex0]->GetType();
99 }
100 }  // namespace
101 
102 MIND_API_OPERATOR_IMPL(SparseSegmentSum, BaseOperator);
SparseSegmentSumInfer(const abstract::AnalysisEnginePtr &,const PrimitivePtr & prim,const std::vector<AbstractBasePtr> & input_args)103 AbstractBasePtr SparseSegmentSumInfer(const abstract::AnalysisEnginePtr &, const PrimitivePtr &prim,
104                                       const std::vector<AbstractBasePtr> &input_args) {
105   MS_EXCEPTION_IF_NULL(prim);
106   auto prim_name = prim->name();
107   const int64_t input_num = static_cast<int64_t>(kInputIndex3);
108   CheckAndConvertUtils::CheckInputArgs(input_args, kEqual, input_num, prim_name);
109   auto types = SparseSegmentSumInferType(prim, input_args);
110   auto shapes = SparseSegmentSumInferShape(prim, input_args);
111   return abstract::MakeAbstract(shapes, types);
112 }
113 
114 // AG means auto generated
115 class MIND_API AGSparseSegmentSumInfer : public abstract::OpInferBase {
116  public:
InferShape(const PrimitivePtr & primitive,const std::vector<AbstractBasePtr> & input_args) const117   BaseShapePtr InferShape(const PrimitivePtr &primitive,
118                           const std::vector<AbstractBasePtr> &input_args) const override {
119     return SparseSegmentSumInferShape(primitive, input_args);
120   }
121 
InferType(const PrimitivePtr & primitive,const std::vector<AbstractBasePtr> & input_args) const122   TypePtr InferType(const PrimitivePtr &primitive, const std::vector<AbstractBasePtr> &input_args) const override {
123     return SparseSegmentSumInferType(primitive, input_args);
124   }
InferShapeAndType(const abstract::AnalysisEnginePtr & engine,const PrimitivePtr & primitive,const std::vector<AbstractBasePtr> & input_args) const125   AbstractBasePtr InferShapeAndType(const abstract::AnalysisEnginePtr &engine, const PrimitivePtr &primitive,
126                                     const std::vector<AbstractBasePtr> &input_args) const override {
127     return SparseSegmentSumInfer(engine, primitive, input_args);
128   }
129 
GetValueDependArgIndices() const130   std::set<int64_t> GetValueDependArgIndices() const override { return {2}; }
131 };
132 
133 REGISTER_PRIMITIVE_OP_INFER_IMPL(SparseSegmentSum, prim::kPrimSparseSegmentSum, AGSparseSegmentSumInfer, false);
134 }  // namespace ops
135 }  // namespace mindspore
136