target_assign_op.h 4.5 KB
Newer Older
1
/* Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved.
2 3 4 5 6 7 8 9 10 11 12 13 14 15

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
Y
Yi Wang 已提交
16 17 18
#include "paddle/fluid/framework/op_registry.h"
#include "paddle/fluid/platform/assert.h"
#include "paddle/fluid/platform/for_range.h"
19 20 21

namespace paddle {
namespace operators {
22
template <typename T, typename WT>
23
struct TargetAssignFunctor {
24
  const T* in_;
25 26
  const int* match_indices_;
  const size_t* lod_;
27 28 29 30 31 32 33 34 35 36 37 38 39 40
  const int mismatch_value_;
  const int64_t N_;
  const int64_t M_;
  const int64_t P_;
  const int64_t K_;

  T* out_;
  WT* out_wt_;

  TargetAssignFunctor(const T* input, const int* match_indices,
                      const size_t* lod, const int mismatch_value,
                      const int64_t N, const int64_t M, const int64_t P,
                      const int64_t K, T* out, WT* out_wt)
      : in_(input),
41 42
        match_indices_(match_indices),
        lod_(lod),
43 44 45 46 47 48 49
        mismatch_value_(mismatch_value),
        N_(N),
        M_(M),
        P_(P),
        K_(K),
        out_(out),
        out_wt_(out_wt) {}
50 51

  HOSTDEVICE void operator()(size_t i) const {
52 53
    int h = i / M_;
    int w = i - h * M_;
54

55 56
    size_t off = lod_[h];
    int id = match_indices_[i];
57

58 59
    T* out = out_ + i * K_;
    WT* out_wt = out_wt_ + i;
60 61

    if (id > -1) {
62 63 64 65 66 67
      int w_off = w % P_;
      const T* in = in_ + ((off + id) * P_ + w_off) * K_;
      for (int64_t k = 0; k < K_; ++k) {
        out[k] = in[k];
      }
      out_wt[0] = static_cast<WT>(1.);
68
    } else {
69 70 71 72
      for (int64_t k = 0; k < K_; ++k) {
        out[k] = static_cast<T>(mismatch_value_);
      }
      out_wt[0] = static_cast<WT>(0.);
73 74 75 76
    }
  }
};

77
template <typename DeviceContext, typename T, typename WT>
D
dangqingqing 已提交
78
struct NegTargetAssignFunctor {
79
  void operator()(const platform::DeviceContext& ctx, const int* neg_indices,
80 81
                  const size_t* lod, const int N, const int M, const int K,
                  const int mismatch_value, T* out, WT* out_wt) const;
82 83
};

84
template <typename DeviceContext, typename T, typename WT>
85 86 87
class TargetAssignKernel : public framework::OpKernel<T> {
 public:
  void Compute(const framework::ExecutionContext& ctx) const override {
88
    auto* x = ctx.Input<framework::LoDTensor>("X");
89 90
    auto* match_indices = ctx.Input<framework::Tensor>("MatchIndices");

91 92
    auto* out = ctx.Output<framework::Tensor>("Out");
    auto* out_wt = ctx.Output<framework::Tensor>("OutWeight");
93

94 95
    PADDLE_ENFORCE_EQ(x->lod().size(), 1UL);
    int mismatch_value = ctx.Attr<int>("mismatch_value");
96

97
    const T* x_data = x->data<T>();
98 99
    const int* match_idx_data = match_indices->data<int>();

100 101
    T* out_data = out->mutable_data<T>(ctx.GetPlace());
    WT* out_wt_data = out_wt->mutable_data<WT>(ctx.GetPlace());
102

103 104 105 106
    int64_t n = match_indices->dims()[0];
    int64_t m = match_indices->dims()[1];
    int64_t p = x->dims()[1];
    int64_t k = x->dims()[2];
107

108
    auto x_lod = x->lod().back();
109
#if defined(PADDLE_WITH_CUDA)
110
    size_t* x_lod_data = x_lod.MutableData(ctx.GetPlace());
111 112 113
#else
    size_t* x_lod_data = x_lod.data();
#endif
114

115 116 117
    TargetAssignFunctor<T, WT> functor(x_data, match_idx_data, x_lod_data,
                                       mismatch_value, n, m, p, k, out_data,
                                       out_wt_data);
118 119

    auto& device_ctx = ctx.template device_context<DeviceContext>();
120
    platform::ForRange<DeviceContext> for_range(device_ctx, n * m);
121 122
    for_range(functor);

123 124 125 126 127
    auto* neg_indices = ctx.Input<framework::LoDTensor>("NegIndices");
    if (neg_indices) {
      PADDLE_ENFORCE_EQ(neg_indices->lod().size(), 1UL);
      const int* neg_idx_data = neg_indices->data<int>();
      auto neg_lod = neg_indices->lod().back();
128
#if defined(PADDLE_WITH_CUDA)
129
      size_t* neg_lod_data = neg_lod.MutableData(ctx.GetPlace());
130 131 132
#else
      size_t* neg_lod_data = neg_lod.data();
#endif
133 134 135 136
      NegTargetAssignFunctor<DeviceContext, T, WT> neg_trg_functor;
      neg_trg_functor(device_ctx, neg_idx_data, neg_lod_data, n, m, k,
                      mismatch_value, out_data, out_wt_data);
    }
137 138 139 140 141
  }
};

}  // namespace operators
}  // namespace paddle