RowConvOp.cpp 8.0 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
/* Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserve.

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 "RowConvOp.h"
D
dangqingqing 已提交
16
#include <iostream>
17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63
#include "paddle/math/Vector.h"

namespace paddle {

template <>
void RowConv<DEVICE_TYPE_CPU>(CpuMatrix& out,
                              const CpuMatrix& in,
                              const CpuMatrix& filter,
                              const CpuIVector& seq) {
  const int* starts = seq.getData();
  const size_t numSeq = seq.getSize() - 1;
  const size_t contextLength = filter.getHeight();
  for (size_t i = 0; i < numSeq; ++i) {
    size_t begin = starts[i];
    size_t end = starts[i + 1];
    for (size_t j = begin; j < end; ++j) {
      MatrixPtr x;
      MatrixPtr w;
      if ((j + contextLength) < end) {
        x = (const_cast<CpuMatrix&>(in)).subMatrix(j, contextLength);
        w = (const_cast<CpuMatrix&>(filter)).subMatrix(0, contextLength);
      } else {
        x = (const_cast<CpuMatrix&>(in)).subMatrix(j, end - j);
        w = (const_cast<CpuMatrix&>(filter)).subMatrix(0, end - j);
      }
      MatrixPtr y = out.subMatrix(j, 1);
      y->addDotMulVMM(*x, *w);
    }
  }
}

template <>
void RowConvGrad<DEVICE_TYPE_CPU>(const CpuMatrix& outG,
                                  const CpuMatrix& in,
                                  const CpuMatrix& filter,
                                  CpuMatrix& inG,
                                  CpuMatrix& filterG,
                                  const CpuIVector& seq) {
  // gradient w.r.t filter
  const int* starts = seq.getData();
  const size_t numSeq = seq.getSize() - 1;
  const size_t contextLength = filter.getHeight();
  if (filterG) {
    for (size_t i = 0; i < numSeq; ++i) {
      size_t begin = starts[i];
      size_t end = starts[i + 1];
      size_t steps = end - begin;
D
dangqingqing 已提交
64
      for (size_t j = 0; j < contextLength && (begin + j) < end; ++j) {
65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83
        MatrixPtr x =
            (const_cast<CpuMatrix&>(in)).subMatrix(begin + j, steps - j);
        MatrixPtr dy =
            (const_cast<CpuMatrix&>(outG)).subMatrix(begin, steps - j);
        MatrixPtr dw = filterG.subMatrix(j, 1);
        dw->addDotMulVMM(*dy, *x);
      }
    }
  }

  // gradient w.r.t input feature
  if (inG) {
    for (size_t i = 0; i < numSeq; ++i) {
      size_t begin = starts[i];
      size_t end = starts[i + 1];
      size_t steps = end - begin;
      for (size_t j = 0; j < steps; ++j) {
        MatrixPtr dx = inG.subMatrix(begin + j, 1);
        for (size_t t = 0; t < contextLength; ++t) {
D
dangqingqing 已提交
84
          if (int(j - t) >= 0) {
85 86 87 88 89 90 91 92 93 94 95 96
            MatrixPtr dy =
                (const_cast<CpuMatrix&>(outG)).subMatrix(begin + j - t, 1);
            MatrixPtr w = (const_cast<CpuMatrix&>(filter)).subMatrix(t, 1);
            dx->addDotMul(*dy, *w, 1.0, 1.0);
          }
        }
      }
    }
  }
}

/**
D
dangqingqing 已提交
97 98 99 100 101 102 103 104
 * \brief The row convolution is called lookahead convolution. It is firstly
 * introduced in deep-speech2 system. The bidirectional RNN that learns
 * representation for a sequence by performing a forward and a backward pass
 * through the entire sequence. However, unlike unidirectional RNNs,
 * bidirectional RNNs are challenging to deploy in an online and low-latency
 * setting. The lookahead convolution incorporates information from future
 * subsequences in a computationally efficient manner to improve unidirectional
 * recurrent neural networks.
105
 *
D
dangqingqing 已提交
106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127
 * The connection of row convolution is different form the 1D sequence
 * convolution. Assumed that, the future context-length is k, that is to say,
 * it can get the output at timestep t by using the the input feature from t-th
 * timestep to (t+k)-th timestep. Assumed that the hidden dim of input
 * activations are d, the activations r_t for the new layer at time-step t are:
 *
 *
 *            -- k + 1
 *  r(t,i) =  >       W(i,j) * h(t+j-1, i),  for (1 <= i <= d)
 *            -- j = 1
 *
 *
 * The weight shape is: (k + 1) x d
 * Function Arguments:
 *
 * \param inputs[0]  The input activations.
 * \param inputs[0]  The filter (or weight) and shape is (k+1) x d.
 * \param outputs[1] The output activations.
 *
 * [1] Dario Amodei, etc. Deep Speech 2 : End-to-End Speech Recognition in
 * English
 *     and Mandarin. https://arxiv.org/abs/1512.02595
128 129 130 131 132 133 134 135 136 137 138
 */

template <DeviceType Device>
class RowConvFunc : public FunctionBase {
public:
  void init(const FuncConfig& config) override {}

  void calc(const BufferArgs& inputs, const BufferArgs& outputs) override {
    // check
    CHECK_EQ(2UL, inputs.size());
    CHECK_EQ(1UL, outputs.size());
D
dangqingqing 已提交
139
    // TODO(qingqing): support ASSIGN_TO.
140 141 142 143 144 145 146 147
    CHECK_EQ(outputs[0].getArgType(), ADD_TO);
    CHECK(inputs[0].isSequenceArg() && outputs[0].isSequenceArg())
        << "SequenceArg required here.";
    const auto in = dynamic_cast<const SequenceArg&>(inputs[0]);
    auto out = dynamic_cast<const SequenceArg&>(outputs[0]);
    auto w = inputs[1];
    CHECK(in.data() && out.data() && in.getSequenceId().data());
    CHECK_EQ(in.shape().ndims(), 2UL);
D
dangqingqing 已提交
148
    CHECK(in.shape() == out.shape());
149 150 151 152 153 154 155 156 157 158
    CHECK_EQ(w.shape()[1], in.shape()[1]);

    auto outMat = out.matrix<Device>();
    const auto inMat = in.matrix<Device>();
    const auto wMat = w.matrix<Device>();
    const auto seqId = in.getSequenceId().vector<int, Device>();

    RowConv<Device>(outMat, inMat, wMat, seqId);
  }
};
D
dangqingqing 已提交
159

160
/**
D
dangqingqing 已提交
161 162
 * \brief The backward of row convolution function. This function calculated
 * the gradient w.r.t filter and the gradient w.r.t input activations(or data).
163 164
 *
 * Argument in this Function:
D
dangqingqing 已提交
165 166 167 168 169 170 171 172 173
 *
 * \param inputs[0]  The gradient w.r.t output activations.
 * \param inputs[1]  The input activations.
 * \param inputs[2]  The filter (or weight) and shape is (k+1) x d.
 * \param outputs[0] The gradient w.r.t input activations.
 * \param outputs[1] The gradient w.r.r filter.
 *
 * Abbreviation:
 * w.r.t: with respect to.
174 175 176 177
 */

template <DeviceType Device>
class RowConvGradFunc : public FunctionBase {
D
dangqingqing 已提交
178
  // TODO(qingqing): split into RowConvDataFunc and RowConvWeightFunc
179 180 181 182
public:
  void init(const FuncConfig& config) override {}

  void calc(const BufferArgs& inputs, const BufferArgs& outputs) override {
D
dangqingqing 已提交
183 184 185 186 187 188 189 190 191
    // check
    CHECK_EQ(3UL, inputs.size());
    CHECK_EQ(2UL, outputs.size());
    CHECK_EQ(outputs[0].getArgType(), ADD_TO);
    CHECK_EQ(outputs[1].getArgType(), ADD_TO);
    CHECK(inputs[0].isSequenceArg() && inputs[1].isSequenceArg() &&
          outputs[0].isSequenceArg())
        << "SequenceArg required here.";

192 193 194 195 196 197
    const auto outGrad = dynamic_cast<const SequenceArg&>(inputs[0]);
    const auto in = dynamic_cast<const SequenceArg&>(inputs[1]);
    const auto w = inputs[2];
    auto inGrad = dynamic_cast<const SequenceArg&>(outputs[0]);
    auto wGrad = outputs[1];

D
dangqingqing 已提交
198
    CHECK_EQ(in.shape().ndims(), 2UL);
D
dangqingqing 已提交
199 200
    CHECK(in.shape() == inGrad.shape());
    CHECK(in.shape() == outGrad.shape());
D
dangqingqing 已提交
201 202
    CHECK_EQ(wGrad.shape()[1], in.shape()[1]);

203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219
    const auto outGMat = outGrad.matrix<Device>();
    const auto inMat = in.matrix<Device>();
    const auto wMat = w.matrix<Device>();
    auto inGMat = inGrad.data()
                      ? inGrad.matrix<Device>()
                      : typename Tensor<real, Device>::Matrix(nullptr, 0, 0);
    auto wGMat = wGrad.data()
                     ? wGrad.matrix<Device>()
                     : typename Tensor<real, Device>::Matrix(nullptr, 0, 0);
    const auto seqId = in.getSequenceId().vector<int, Device>();

    RowConvGrad<Device>(outGMat, inMat, wMat, inGMat, wGMat, seqId);
  }
};

REGISTER_TYPED_FUNC(RowConv, CPU, RowConvFunc);
REGISTER_TYPED_FUNC(RowConvGrad, CPU, RowConvGradFunc);
220
#ifdef PADDLE_WITH_GPU
221
REGISTER_TYPED_FUNC(RowConv, GPU, RowConvFunc);
D
dangqingqing 已提交
222
REGISTER_TYPED_FUNC(RowConvGrad, GPU, RowConvGradFunc);
223 224 225
#endif

}  // namespace paddle