sum_op.cc 8.5 KB
Newer Older
1
/* Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserved.
2 3 4 5 6 7 8 9 10 11
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. */

Y
Yi Wang 已提交
12
#include "paddle/fluid/operators/sum_op.h"
13

14 15
#include <algorithm>
#include <string>
16
#include <vector>
17

Y
Yi Wang 已提交
18 19
#include "paddle/fluid/framework/var_type_inference.h"
#include "paddle/fluid/operators/detail/safe_ref.h"
20

21 22 23 24
#ifdef PADDLE_WITH_MKLDNN
#include "paddle/fluid/platform/mkldnn_helper.h"
#endif

25 26 27 28 29 30 31 32
namespace paddle {
namespace operators {
using framework::Tensor;

class SumOp : public framework::OperatorWithKernel {
 public:
  using framework::OperatorWithKernel::OperatorWithKernel;

33
  void InferShape(framework::InferShapeContext* ctx) const override {
Q
qiaolongfei 已提交
34
    PADDLE_ENFORCE(ctx->HasInputs("X"), "Inputs(X) should not be null");
35

Q
Qiao Longfei 已提交
36 37
    PADDLE_ENFORCE(ctx->HasOutput("Out"),
                   "Output(Out) of SumOp should not be null.");
38 39
    if (ctx->IsRuntime() &&
        ctx->GetOutputsVarType("Out")[0] ==
40
            framework::proto::VarType::LOD_TENSOR_ARRAY) {
41 42
      return;  // skip runtime infershape when is tensor array;
    }
43

44
    auto x_dims = ctx->GetInputsDim("X");
Q
Qiao Longfei 已提交
45
    size_t N = x_dims.size();
46 47
    PADDLE_ENFORCE_GT(N, 0, "Input tensors count should > 0.");
    if (N == 1) {
M
minqiyang 已提交
48
      VLOG(3) << "Warning: sum have only one input, may waste memory";
49
    }
Q
qiaolongfei 已提交
50

51 52 53 54 55 56 57 58 59 60
    framework::DDim in_dim({0});
    for (auto& x_dim : x_dims) {
      if (framework::product(x_dim) == 0) {
        continue;
      }
      if (framework::product(in_dim) == 0) {
        in_dim = x_dim;
      } else {
        PADDLE_ENFORCE_EQ(in_dim, x_dim, "Input tensors must have same shape");
      }
Q
qijun 已提交
61
    }
Q
Qiao Longfei 已提交
62 63
    ctx->SetOutputDim("Out", in_dim);
    ctx->ShareLoD("X", /*->*/ "Out");
64
  }
65 66

 protected:
67
  framework::OpKernelType GetExpectedKernelType(
68 69
      const framework::ExecutionContext& ctx) const override {
    auto x_vars = ctx.MultiInputVar("X");
C
chengduo 已提交
70
    auto x_vars_name = ctx.Inputs("X");
71 72 73 74 75 76 77 78 79 80 81 82

    framework::LibraryType library{framework::LibraryType::kPlain};
    framework::DataLayout layout{framework::DataLayout::kAnyLayout};

#ifdef PADDLE_WITH_MKLDNN
    if (library == framework::LibraryType::kPlain &&
        platform::CanMKLDNNBeUsed(ctx)) {
      library = framework::LibraryType::kMKLDNN;
      layout = framework::DataLayout::kMKLDNN;
    }
#endif

83
    if (x_vars[0]->IsType<framework::LoDTensor>()) {
84
      int dtype = -1;
C
chengduo 已提交
85 86 87
      for (size_t idx = 0; idx < x_vars.size(); ++idx) {
        PADDLE_ENFORCE(x_vars[idx] != nullptr,
                       "Input var[%s] should not be nullptr", x_vars_name[idx]);
C
chengduo 已提交
88 89
        auto tensor =
            framework::GetLoDTensorOrSelectedRowsValueFromVar(*x_vars[idx]);
90
        if (tensor->numel() == 0) {
91 92 93
          continue;
        }
        if (dtype == -1) {
M
minqiyang 已提交
94
          dtype = tensor->type();
95
        } else {
M
minqiyang 已提交
96
          PADDLE_ENFORCE_EQ(dtype, tensor->type());
97 98 99 100 101
        }
      }
      PADDLE_ENFORCE_NE(dtype, -1,
                        "Sum operator should have at least one tensor");

102
      return framework::OpKernelType(
103 104
          static_cast<framework::proto::VarType::Type>(dtype), ctx.GetPlace(),
          layout, library);
105
    } else if (x_vars[0]->IsType<framework::SelectedRows>()) {
106 107 108
      for (auto& var : x_vars) {
        auto& value = var->Get<framework::SelectedRows>().value();
        if (value.IsInitialized()) {
M
minqiyang 已提交
109 110
          return framework::OpKernelType(value.type(), ctx.device_context(),
                                         layout, library);
111 112 113 114
        }
      }
      // if input sparse vars are not initialized, use an default kernel type.
      return framework::OpKernelType(framework::proto::VarType::FP32,
115
                                     ctx.device_context(), layout, library);
116
    } else if (x_vars[0]->IsType<framework::LoDTensorArray>()) {
Y
Yang Yang(Tony) 已提交
117 118 119 120
      for (auto& x_var : x_vars) {
        auto& array = x_var->Get<framework::LoDTensorArray>();
        for (auto& each : array) {
          if (each.numel() != 0) {
M
minqiyang 已提交
121 122
            return framework::OpKernelType(each.type(), ctx.device_context(),
                                           layout, library);
Y
Yang Yang(Tony) 已提交
123
          }
124 125
        }
      }
Y
Yang Yang(Tony) 已提交
126
      PADDLE_THROW("Cannot find the input data type by all input data");
127 128 129 130
    }
    PADDLE_THROW("Unexpected branch. Input type is %s",
                 x_vars[0]->Type().name());
  }
131 132 133 134
};

class SumOpMaker : public framework::OpProtoAndCheckerMaker {
 public:
Y
Yu Yang 已提交
135
  void Make() override {
136 137
    AddInput("X", "(vector<Tensor>) The input tensors of sum operator.")
        .AsDuplicable();
138
    AddOutput("Out", "(Tensor) The output tensor of sum operator.");
139 140 141
    AddAttr<bool>("use_mkldnn",
                  "(bool, default false) Only used in mkldnn kernel")
        .SetDefault(false);
142
    AddComment(R"DOC(
143
Sum operator.
144

145 146
This operators sums the input tensors. All the inputs can carry the
LoD (Level of Details) information. However, the output only shares
147
the LoD information with the first input.
148
)DOC");
149 150 151
  }
};

Q
QI JUN 已提交
152 153
class SumOpVarTypeInference : public framework::VarTypeInference {
 public:
Y
Yu Yang 已提交
154 155
  void operator()(const framework::OpDesc& op_desc,
                  framework::BlockDesc* block) const override {
Q
QI JUN 已提交
156
    auto& inputs = op_desc.Input("X");
157
    auto var_type = framework::proto::VarType::SELECTED_ROWS;
Y
Yang Yang(Tony) 已提交
158
    for (auto& name : op_desc.Input("X")) {
M
minqiyang 已提交
159 160
      VLOG(10) << name << " "
               << block->FindRecursiveOrCreateVar(name).GetType();
Y
Yang Yang(Tony) 已提交
161 162
    }

Q
QI JUN 已提交
163 164
    bool any_input_is_lod_tensor = std::any_of(
        inputs.begin(), inputs.end(), [block](const std::string& name) {
Y
Yang Yu 已提交
165
          return block->FindRecursiveOrCreateVar(name).GetType() ==
166
                 framework::proto::VarType::LOD_TENSOR;
Q
QI JUN 已提交
167
        });
168 169

    auto is_tensor_array = [block](const std::string& name) {
Y
Yang Yu 已提交
170
      return block->FindRecursiveOrCreateVar(name).GetType() ==
171
             framework::proto::VarType::LOD_TENSOR_ARRAY;
172 173 174 175 176 177 178 179
    };

    bool any_input_is_tensor_array =
        std::any_of(inputs.begin(), inputs.end(), is_tensor_array);
    bool all_inputs_are_tensor_array =
        std::all_of(inputs.begin(), inputs.end(), is_tensor_array);

    if (any_input_is_tensor_array) {
Y
Yang Yang(Tony) 已提交
180 181 182 183
      if (!all_inputs_are_tensor_array) {
        std::ostringstream os;
        for (auto& each : inputs) {
          os << "    " << each << " type is "
Y
Yang Yu 已提交
184
             << block->FindRecursiveOrCreateVar(each).GetType() << "\n";
Y
Yang Yang(Tony) 已提交
185 186 187 188
        }
        PADDLE_ENFORCE(all_inputs_are_tensor_array,
                       "Not all inputs are tensor array:\n%s", os.str());
      }
189
      var_type = framework::proto::VarType::LOD_TENSOR_ARRAY;
190
    } else if (any_input_is_lod_tensor) {
191
      var_type = framework::proto::VarType::LOD_TENSOR;
Q
QI JUN 已提交
192 193 194
    }

    auto out_var_name = op_desc.Output("Out").front();
Y
Yang Yu 已提交
195
    auto& out_var = block->FindRecursiveOrCreateVar(out_var_name);
Y
Yang Yang(Tony) 已提交
196 197 198
    out_var.SetType(var_type);
    auto& in_var = detail::Ref(block->FindVarRecursive(inputs.front()));
    out_var.SetDataType(in_var.GetDataType());
Q
QI JUN 已提交
199 200 201
  }
};

202
class SumGradMaker : public framework::GradOpDescMakerBase {
203
 public:
204
  using framework::GradOpDescMakerBase::GradOpDescMakerBase;
205

Y
Yu Yang 已提交
206
  std::vector<std::unique_ptr<framework::OpDesc>> operator()() const override {
207
    auto x_grads = InputGrad("X", false);
Y
Yu Yang 已提交
208
    std::vector<std::unique_ptr<framework::OpDesc>> grad_ops;
209 210 211 212
    grad_ops.reserve(x_grads.size());
    auto og = OutputGrad("Out");
    std::transform(x_grads.begin(), x_grads.end(), std::back_inserter(grad_ops),
                   [&og](const std::string& x_grad) {
Y
Yu Yang 已提交
213
                     auto* grad_op = new framework::OpDesc();
Y
Yu Yang 已提交
214 215 216 217
                     grad_op->SetType("scale");
                     grad_op->SetInput("X", og);
                     grad_op->SetOutput("Out", {x_grad});
                     grad_op->SetAttr("scale", 1.0f);
Y
Yu Yang 已提交
218
                     return std::unique_ptr<framework::OpDesc>(grad_op);
219 220
                   });
    return grad_ops;
221 222 223 224 225 226 227
  }
};

}  // namespace operators
}  // namespace paddle

namespace ops = paddle::operators;
228

Q
QI JUN 已提交
229 230
REGISTER_OPERATOR(sum, ops::SumOp, ops::SumOpMaker, ops::SumGradMaker,
                  ops::SumOpVarTypeInference);
231

Q
QI JUN 已提交
232 233 234 235 236
REGISTER_OP_CPU_KERNEL(
    sum, ops::SumKernel<paddle::platform::CPUDeviceContext, float>,
    ops::SumKernel<paddle::platform::CPUDeviceContext, double>,
    ops::SumKernel<paddle::platform::CPUDeviceContext, int>,
    ops::SumKernel<paddle::platform::CPUDeviceContext, int64_t>);