conv_transpose_op.h 12.0 KB
Newer Older
C
chengduoZH 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
/* 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. */

#pragma once

#include "paddle/framework/eigen.h"
#include "paddle/framework/op_registry.h"
C
chengduoZH 已提交
19
#include "paddle/operators/math/im2col.h"
C
chengduoZH 已提交
20 21 22 23 24 25 26 27 28 29 30
#include "paddle/operators/math/math_function.h"
#include "paddle/operators/math/vol2col.h"

namespace paddle {
namespace operators {

using Tensor = framework::Tensor;
using DDim = framework::DDim;

// Define Op classes in .h file so that other conv transpose
// operator implementations can reuse the code.
C
chengduoZH 已提交
31 32 33 34 35 36
class Conv2DTransposeOpMaker : public framework::OpProtoAndCheckerMaker {
 public:
  Conv2DTransposeOpMaker(framework::OpProto* proto,
                         framework::OpAttrChecker* op_checker);
};

C
chengduoZH 已提交
37 38 39 40 41 42
class Conv3DTransposeOpMaker : public framework::OpProtoAndCheckerMaker {
 public:
  Conv3DTransposeOpMaker(framework::OpProto* proto,
                         framework::OpAttrChecker* op_checker);
};

C
chengduoZH 已提交
43
class ConvTransposeOp : public framework::OperatorWithKernel {
C
chengduoZH 已提交
44 45 46 47 48 49 50
 public:
  using framework::OperatorWithKernel::OperatorWithKernel;

 protected:
  void InferShape(framework::InferShapeContext* ctx) const override;
};

C
chengduoZH 已提交
51
class ConvTransposeOpGrad : public framework::OperatorWithKernel {
C
chengduoZH 已提交
52 53 54 55 56 57 58 59
 public:
  using framework::OperatorWithKernel::OperatorWithKernel;

 protected:
  void InferShape(framework::InferShapeContext* ctx) const override;
};

template <typename Place, typename T>
60
class GemmConvTransposeKernel : public framework::OpKernel<T> {
C
chengduoZH 已提交
61 62 63 64 65 66 67 68
 public:
  void Compute(const framework::ExecutionContext& context) const override {
    const Tensor* input = context.Input<Tensor>("Input");
    // The filter will be reshaped, so it should not be constant pointer
    Tensor filter = *context.Input<Tensor>("Filter");
    Tensor* output = context.Output<Tensor>("Output");

    std::vector<int> strides = context.Attr<std::vector<int>>("strides");
C
chengduoZH 已提交
69 70 71
    // TODO(Zhuoyuan): Paddings can be added in future.
    // groups will alway be disabled in conv2dtranspose.

C
chengduoZH 已提交
72
    const int batch_size = static_cast<int>(input->dims()[0]);
C
chengduoZH 已提交
73

74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92
    // input_shape_vec: {h, w} or {d, h, w}
    std::vector<int64_t> input_shape_vec = framework::vectorize(input->dims());
    input_shape_vec.erase(input_shape_vec.begin(), input_shape_vec.begin() + 2);

    // filter_shape_vec: {k_h, k_w} or {k_d, k_h, k_w}
    std::vector<int64_t> filter_shape_vec = framework::vectorize(filter.dims());
    filter_shape_vec.erase(filter_shape_vec.begin(),
                           filter_shape_vec.begin() + 2);

    // use col_shape in the im2col and col2im (or vol2col and col2vol)
    // calculation
    // col_shape_vec: {c, k_h, k_w, h, w} or {c, k_d, k_h, k_w, d, h, w}
    std::vector<int64_t> col_shape_vec;
    col_shape_vec.push_back(output->dims()[1]);
    col_shape_vec.insert(col_shape_vec.end(), filter_shape_vec.begin(),
                         filter_shape_vec.end());
    col_shape_vec.insert(col_shape_vec.end(), input_shape_vec.begin(),
                         input_shape_vec.end());
    DDim col_shape(framework::make_ddim(col_shape_vec));
C
chengduoZH 已提交
93 94

    // use col_matrix_shape in the gemm calculation
95 96 97
    // size: (c * k_h * k_w, h * w) or (c * k_d * k_h * k_w, d * h * w)
    DDim col_matrix_shape =
        framework::flatten_to_2d(col_shape, filter_shape_vec.size() + 1);
C
chengduoZH 已提交
98 99 100 101 102 103 104 105 106 107

    Tensor col;
    col.mutable_data<T>(col_shape, context.GetPlace());
    // col_matrix shares the same piece of data with col,
    // but will be reshaped into a two-dimensional matrix shape
    // to call the matrix multiplication interface.
    Tensor col_matrix;
    col_matrix.ShareDataWith(col);
    col_matrix.Resize(col_matrix_shape);

108 109 110
    // output size: (c, o_h, o_w) or (c, o_d, o_h, o_w)
    DDim output_shape =
        framework::slice_ddim(output->dims(), 1, output->dims().size());
C
chengduoZH 已提交
111

112 113 114 115 116
    // input matrix size: (m, h * w) or (m, d * h * w)
    DDim input_matrix_shape = {input->dims()[1], col_matrix_shape[1]};

    // filter size: (m, c * k_h * k_w) or (m, c * k_d * k_h * k_w)
    DDim filter_matrix_shape = {input->dims()[1], col_matrix_shape[0]};
C
chengduoZH 已提交
117 118 119
    filter.Resize(filter_matrix_shape);

    output->mutable_data<T>(context.GetPlace());
C
chengduoZH 已提交
120 121
    math::SetConstant<Place, T> set_zero;
    set_zero(context.device_context(), output, static_cast<T>(0));
C
chengduoZH 已提交
122

123 124
    // convolution transpose: gemm + col2im or col2vol (similar to conv-backward
    // on input)
C
chengduoZH 已提交
125
    for (int i = 0; i < batch_size; i++) {
126
      // batch with size (m, h * w) or (m, d * h * w)
C
chengduoZH 已提交
127 128
      Tensor input_batch = input->Slice(i, i + 1).Resize(input_matrix_shape);

129
      // output size: (c, o_h, o_w) or (c, o_d, o_h, o_w)
C
chengduoZH 已提交
130 131 132
      Tensor output_batch = output->Slice(i, i + 1).Resize(output_shape);

      // col_matrix = filter * input_batch
133
      // of shape (c * k_h * k_w, h * w) or (c * k_d * k_h * k_w, d * h * w)
C
chengduoZH 已提交
134
      math::matmul<Place, T>(context.device_context(), filter, true,
C
chengduoZH 已提交
135 136 137
                             input_batch, false, static_cast<T>(1.0),
                             &col_matrix, static_cast<T>(0.0));

138 139 140 141 142 143 144 145 146 147 148 149 150 151
      if (filter_shape_vec.size() == 2) {
        // col2im: col_matrix -> dy
        // from (c * k_h * k_w, h * w) to (c, o_h, o_w)
        math::Col2ImFunctor<math::ColFormat::kCFO, Place, T> col2im;

        col2im(context.device_context(), output_batch, col, strides[0],
               strides[1], 0, 0, 0, 0);
      } else if (filter_shape_vec.size() == 3) {
        // col2vol: col_matrix -> dy
        // from (c * k_d * k_h * k_w, d * h * w) to (c, o_d, o_h, o_w)
        math::Col2VolFunctor<Place, T> col2vol;
        col2vol(context.device_context(), output_batch, col, strides[0],
                strides[1], strides[2], 0, 0, 0);
      }
C
chengduoZH 已提交
152 153 154 155 156
    }
  }
};

template <typename Place, typename T>
157
class GemmConvTransposeGradKernel : public framework::OpKernel<T> {
C
chengduoZH 已提交
158 159 160 161 162 163 164 165 166 167 168 169 170
 public:
  void Compute(const framework::ExecutionContext& context) const override {
    const Tensor* input = context.Input<Tensor>("Input");
    const Tensor* output_grad =
        context.Input<Tensor>(framework::GradVarName("Output"));
    // For filter, we do not use const pointer b/c we will do reshape,
    // but we should avoid modifying its value.
    Tensor filter = *context.Input<Tensor>("Filter");
    Tensor* input_grad =
        context.Output<Tensor>(framework::GradVarName("Input"));
    Tensor* filter_grad =
        context.Output<Tensor>(framework::GradVarName("Filter"));

171 172
    if ((!input_grad) && (!filter_grad)) return;

C
chengduoZH 已提交
173 174 175 176
    std::vector<int> strides = context.Attr<std::vector<int>>("strides");
    // Actually, no paddings and groups allowed in conv transpose.
    std::vector<int> paddings = context.Attr<std::vector<int>>("paddings");

C
chengduoZH 已提交
177
    const int batch_size = static_cast<int>(input->dims()[0]);
C
chengduoZH 已提交
178

179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197
    // input_shape_vec: {h, w} or {d, h, w}
    std::vector<int64_t> input_shape_vec = framework::vectorize(input->dims());
    input_shape_vec.erase(input_shape_vec.begin(), input_shape_vec.begin() + 2);

    // filter_shape_vec: {k_h, k_w} or {k_d, k_h, k_w}
    std::vector<int64_t> filter_shape_vec = framework::vectorize(filter.dims());
    filter_shape_vec.erase(filter_shape_vec.begin(),
                           filter_shape_vec.begin() + 2);

    // use col_shape in the im2col and col2im (or vol2col and col2vol)
    // calculation
    // col_shape_vec: {c, k_h, k_w, h, w} or {c, k_d, k_h, k_w, d, h, w}
    std::vector<int64_t> col_shape_vec;
    col_shape_vec.push_back(output_grad->dims()[1]);
    col_shape_vec.insert(col_shape_vec.end(), filter_shape_vec.begin(),
                         filter_shape_vec.end());
    col_shape_vec.insert(col_shape_vec.end(), input_shape_vec.begin(),
                         input_shape_vec.end());
    DDim col_shape(framework::make_ddim(col_shape_vec));
C
chengduoZH 已提交
198

199 200 201 202
    // use col_matrix_shape in the gemm calculation
    // size: (c * k_h * k_w, h * w) or (c * k_d * k_h * k_w, d * h * w)
    DDim col_matrix_shape =
        framework::flatten_to_2d(col_shape, filter_shape_vec.size() + 1);
C
chengduoZH 已提交
203

204 205 206
    // output size: (c, o_h, o_w) or (c, o_d, o_h, o_w)
    DDim output_shape = framework::slice_ddim(output_grad->dims(), 1,
                                              output_grad->dims().size());
C
chengduoZH 已提交
207

208 209
    // input matrix size: (m, h * w) or (m, d * h * w)
    DDim input_matrix_shape = {input->dims()[1], col_matrix_shape[1]};
C
chengduoZH 已提交
210

211 212
    // filter size: (m, c * k_h * k_w) or (m, c * k_d * k_h * k_w)
    DDim filter_matrix_shape = {input->dims()[1], col_matrix_shape[0]};
C
chengduoZH 已提交
213 214 215 216 217
    filter.Resize(filter_matrix_shape);

    // convolution transpose grad on input:
    // im2col + gemm (similar to conv-forward)
    // input need to compute gradient
C
chengduoZH 已提交
218 219 220 221 222 223
    if (input_grad || filter_grad) {
      Tensor col;
      col.mutable_data<T>(col_shape, context.GetPlace());
      // col_matrix shares the same piece of data with col,
      // but will be reshaped into a two-dimensional matrix shape
      // to call the matrix multiplication interface.
C
chengduoZH 已提交
224 225 226 227
      Tensor col_matrix;
      col_matrix.ShareDataWith(col);
      col_matrix.Resize(col_matrix_shape);

C
chengduoZH 已提交
228 229
      Tensor filter_grad_;
      math::SetConstant<Place, T> set_zero;
C
chengduoZH 已提交
230

C
chengduoZH 已提交
231 232 233 234 235 236 237 238 239
      if (input_grad) {
        input_grad->mutable_data<T>(context.GetPlace());
        set_zero(context.device_context(), input_grad, static_cast<T>(0));
      }
      if (filter_grad) {  // filter size (m, c, k_h, k_w)
        filter_grad->mutable_data<T>(context.GetPlace());
        set_zero(context.device_context(), filter_grad, static_cast<T>(0));
        filter_grad_ = *filter_grad;
        filter_grad_.Resize(filter_matrix_shape);
C
chengduoZH 已提交
240 241
      }

C
chengduoZH 已提交
242 243
      for (int i = 0; i < batch_size; i++) {
        // batch with size (c, o_h * o_w)
C
chengduoZH 已提交
244 245 246
        Tensor output_grad_batch =
            output_grad->Slice(i, i + 1).Resize(output_shape);

247 248 249 250 251 252 253 254 255 256 257 258 259 260 261
        if (filter_shape_vec.size() == 2) {
          // im2col: dy -> col matrix
          // from (c, o_h, o_w) to (c * k_h * k_w, h * w)
          math::Im2ColFunctor<math::ColFormat::kCFO, Place, T> im2col;
          im2col(context.device_context(), output_grad_batch, col, strides[0],
                 strides[1], paddings[0], paddings[0], paddings[1],
                 paddings[1]);
        } else if (filter_shape_vec.size() == 3) {
          // vol2col: dy -> col_matrix
          // from (c, o_d, o_h, o_w) to (c * k_d * k_h * k_w, d * h * w)
          math::Vol2ColFunctor<Place, T> vol2col;
          vol2col(context.device_context(), output_grad_batch, col, strides[0],
                  strides[1], strides[2], paddings[0], paddings[1],
                  paddings[2]);
        }
C
chengduoZH 已提交
262

C
chengduoZH 已提交
263 264 265 266 267 268
        if (input_grad) {
          // batch with size (m, h, w)
          Tensor input_grad_batch =
              input_grad->Slice(i, i + 1).Resize(input_matrix_shape);
          // gemm: dx = filter * dy
          // (m, c * k_h * k_w) * (c * k_h * k_w, h * w) -> (m, h * w)
269
          // or
C
chengduoZH 已提交
270 271 272 273 274 275 276 277 278 279
          // (m, c * k_d * k_h * k_w) * (c * k_d * k_h * k_w, d * h * w) -> (m,
          // d, h, w)
          math::matmul<Place, T>(context.device_context(), filter, false,
                                 col_matrix, false, static_cast<T>(1.0),
                                 &input_grad_batch, static_cast<T>(0.0));
        }
        if (filter_grad) {
          // input batch
          Tensor in_batch = input->Slice(i, i + 1).Resize(input_matrix_shape);
          // gemm: d_filter = x * dy^T
280 281
          // (m, c * h * w) * (k_h * k_w, c * h * w) -> (m, k_h * k_w)
          // or
C
chengduoZH 已提交
282 283 284 285 286 287
          // (m, d * h * w) * (d * h * w, c * k_d * k_h * k_w) -> (m, c * k_d *
          // k_h * k_w)
          math::matmul<Place, T>(context.device_context(), in_batch, false,
                                 col_matrix, true, static_cast<T>(1.0),
                                 &filter_grad_, static_cast<T>(1.0));
        }
C
chengduoZH 已提交
288 289 290 291 292 293
      }
    }
  }
};
}  // namespace operators
}  // namespace paddle
反馈
建议
客服 返回
顶部