sum_op.h 6.3 KB
Newer Older
1
/* Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserved.
2 3 4 5 6 7 8 9 10 11 12
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. */

#pragma once
13
#include <vector>
Y
Yi Wang 已提交
14 15 16 17 18
#include "paddle/fluid/framework/eigen.h"
#include "paddle/fluid/framework/lod_tensor_array.h"
#include "paddle/fluid/framework/op_registry.h"
#include "paddle/fluid/operators/math/math_function.h"
#include "paddle/fluid/operators/math/selected_rows_functor.h"
19 20 21 22 23

namespace paddle {
namespace operators {

using Tensor = framework::Tensor;
Q
QI JUN 已提交
24 25
using SelectedRows = framework::SelectedRows;
using LoDTensor = framework::LoDTensor;
26 27 28 29
template <typename T, int MajorType = Eigen::RowMajor,
          typename IndexType = Eigen::DenseIndex>
using EigenVector = framework::EigenVector<T, MajorType, IndexType>;

Q
QI JUN 已提交
30
template <typename DeviceContext, typename T>
Y
Yu Yang 已提交
31
class SumKernel : public framework::OpKernel<T> {
32
 public:
33
  void Compute(const framework::ExecutionContext &context) const override {
Y
Yu Yang 已提交
34
    auto in_vars = context.MultiInputVar("X");
Q
QI JUN 已提交
35 36 37
    int N = in_vars.size();
    auto out_var = context.OutputVar("Out");

Y
Yu Yang 已提交
38 39
    bool in_place = out_var == in_vars[0];

Q
QI JUN 已提交
40
    if (out_var->IsType<framework::LoDTensor>()) {
Y
Update  
Yang Yu 已提交
41
      auto *out = context.Output<LoDTensor>("Out");
Y
Yu Yang 已提交
42
      if (!in_place) {
Y
Refine  
Yang Yu 已提交
43
        out->mutable_data<T>(context.GetPlace());
Y
Update  
Yang Yu 已提交
44 45 46
      }
      auto result = EigenVector<T>::Flatten(*out);
      if (!in_place) {
Q
QI JUN 已提交
47 48
        math::SetConstant<DeviceContext, T> constant_functor;
        constant_functor(context.template device_context<DeviceContext>(), out,
49
                         0.0);
Y
Yu Yang 已提交
50
      }
Q
QI JUN 已提交
51

Q
QI JUN 已提交
52 53 54
      math::SelectedRowsAddToTensor<DeviceContext, T> functor;
      auto &place =
          *context.template device_context<DeviceContext>().eigen_device();
Y
Yu Yang 已提交
55 56
      // If in_place, just skip the first tensor
      for (int i = in_place ? 1 : 0; i < N; i++) {
Q
QI JUN 已提交
57
        if (in_vars[i]->IsType<framework::LoDTensor>()) {
58
          auto &in_t = in_vars[i]->Get<framework::LoDTensor>();
59 60 61
          if (in_t.numel() == 0) {
            continue;
          }
Q
QI JUN 已提交
62 63 64
          auto in = EigenVector<T>::Flatten(in_t);
          result.device(place) = result + in;
        } else if (in_vars[i]->IsType<framework::SelectedRows>()) {
65
          auto &in_t = in_vars[i]->Get<framework::SelectedRows>();
Q
QI JUN 已提交
66
          functor(context.template device_context<DeviceContext>(), in_t, out);
Q
QI JUN 已提交
67 68 69 70 71
        } else {
          PADDLE_THROW("Variable type must be LoDTensor/SelectedRows.");
        }
      }
    } else if (out_var->IsType<framework::SelectedRows>()) {
Y
Yang Yu 已提交
72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97
      std::unique_ptr<framework::SelectedRows> in0;
      if (in_place) {
        // If is in_place, we store the input[0] to in0
        auto &in_sel0 = in_vars[0]->Get<SelectedRows>();
        auto &rows = in_sel0.rows();
#ifdef PADDLE_WITH_CUDA
        std::vector<int64_t> rows_in_cpu;
        rows_in_cpu.reserve(rows.size());
        for (auto item : rows) {
          rows_in_cpu.push_back(item);
        }
        in0.reset(new framework::SelectedRows(rows_in_cpu, in_sel0.height()));
#else
        in0.reset(new framework::SelectedRows(rows, in_sel0.height()));
#endif
        in0->mutable_value()->ShareDataWith(in_sel0.value());
      }

      auto get_selected_row = [&](size_t i) -> const SelectedRows & {
        if (i == 0 && in0) {
          return *in0.get();
        } else {
          return in_vars[i]->Get<SelectedRows>();
        }
      };

98
      auto *out = context.Output<SelectedRows>("Out");
Y
Yancey 已提交
99
      out->mutable_rows()->clear();
100
      auto *out_value = out->mutable_value();
Q
QI JUN 已提交
101 102 103 104

      // Runtime InferShape
      size_t first_dim = 0;
      for (int i = 0; i < N; i++) {
Y
Yang Yu 已提交
105 106
        auto &sel_row = get_selected_row(i);
        first_dim += sel_row.rows().size();
Q
QI JUN 已提交
107
      }
Y
Yang Yu 已提交
108 109 110
      auto in_dim =
          framework::vectorize(get_selected_row(N - 1).value().dims());
      in_dim[0] = static_cast<int64_t>(first_dim);
Q
QI JUN 已提交
111

Y
Yang Yu 已提交
112
      out_value->Resize(framework::make_ddim(in_dim));
113 114 115 116 117 118

      // if all the input sparse vars are empty, no need to
      // merge these vars.
      if (first_dim == 0UL) {
        return;
      }
Q
QI JUN 已提交
119 120
      out_value->mutable_data<T>(context.GetPlace());

Q
QI JUN 已提交
121
      math::SelectedRowsAddTo<DeviceContext, T> functor;
Q
QI JUN 已提交
122 123 124

      int64_t offset = 0;
      for (int i = 0; i < N; i++) {
Y
Yang Yu 已提交
125
        auto &sel_row = get_selected_row(i);
126
        if (sel_row.rows().size() == 0) {
127 128
          continue;
        }
Y
Yang Yu 已提交
129 130 131 132
        PADDLE_ENFORCE_EQ(out->height(), sel_row.height());
        functor(context.template device_context<DeviceContext>(), sel_row,
                offset, out);
        offset += sel_row.value().numel();
Q
QI JUN 已提交
133
      }
134 135 136 137 138 139 140 141 142 143 144 145 146
    } else if (out_var->IsType<framework::LoDTensorArray>()) {
      auto &out_array = *out_var->GetMutable<framework::LoDTensorArray>();
      for (size_t i = in_place ? 1 : 0; i < in_vars.size(); ++i) {
        PADDLE_ENFORCE(in_vars[i]->IsType<framework::LoDTensorArray>(),
                       "Only support all inputs are TensorArray");
        auto &in_array = in_vars[i]->Get<framework::LoDTensorArray>();

        for (size_t i = 0; i < in_array.size(); ++i) {
          if (in_array[i].numel() != 0) {
            if (i >= out_array.size()) {
              out_array.resize(i + 1);
            }
            if (out_array[i].numel() == 0) {
Y
Yi Wang 已提交
147 148
              framework::TensorCopy(in_array[i], in_array[i].place(),
                                    context.device_context(), &out_array[i]);
149 150 151 152 153
              out_array[i].set_lod(in_array[i].lod());
            } else {
              PADDLE_ENFORCE(out_array[i].lod() == in_array[i].lod());
              auto in = EigenVector<T>::Flatten(in_array[i]);
              auto result = EigenVector<T>::Flatten(out_array[i]);
Q
QI JUN 已提交
154 155
              result.device(*context.template device_context<DeviceContext>()
                                 .eigen_device()) = result + in;
156 157 158 159 160 161 162
            }
          }
        }
      }
    } else {
      PADDLE_THROW("Unexpected branch, output variable type is %s",
                   out_var->Type().name());
163 164 165 166 167
    }
  }
};
}  // namespace operators
}  // namespace paddle