conv_op.cc 17.0 KB
Newer Older
1
/* Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserved.
C
chengduoZH 已提交
2

L
Luo Tao 已提交
3 4 5
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
C
chengduoZH 已提交
6

L
Luo Tao 已提交
7
    http://www.apache.org/licenses/LICENSE-2.0
C
chengduoZH 已提交
8

L
Luo Tao 已提交
9 10 11 12 13
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. */
C
chengduoZH 已提交
14

Y
Yi Wang 已提交
15
#include "paddle/fluid/operators/conv_op.h"
Y
Update  
Yi Wang 已提交
16 17 18 19

#include <string>
#include <vector>

20 21 22 23 24 25
#ifdef PADDLE_WITH_CUDA
#include "paddle/fluid/platform/cudnn_helper.h"
#endif
#ifdef PADDLE_WITH_MKLDNN
#include "paddle/fluid/platform/mkldnn_helper.h"
#endif
C
chengduoZH 已提交
26 27 28 29

namespace paddle {
namespace operators {

C
chengduoZH 已提交
30
void ConvOp::InferShape(framework::InferShapeContext* ctx) const {
C
chengduoZH 已提交
31
  PADDLE_ENFORCE(ctx->HasInput("Input"),
C
chengduoZH 已提交
32
                 "Input(Input) of ConvOp should not be null.");
C
chengduoZH 已提交
33
  PADDLE_ENFORCE(ctx->HasInput("Filter"),
C
chengduoZH 已提交
34
                 "Input(Filter) of ConvOp should not be null.");
C
chengduoZH 已提交
35
  PADDLE_ENFORCE(ctx->HasOutput("Output"),
C
chengduoZH 已提交
36
                 "Output(Output) of ConvOp should not be null.");
C
chengduoZH 已提交
37 38 39

  auto in_dims = ctx->GetInputDim("Input");
  auto filter_dims = ctx->GetInputDim("Filter");
40

C
chengduoZH 已提交
41 42 43
  std::vector<int> strides = ctx->Attrs().Get<std::vector<int>>("strides");
  std::vector<int> paddings = ctx->Attrs().Get<std::vector<int>>("paddings");
  int groups = ctx->Attrs().Get<int>("groups");
C
chengduoZH 已提交
44
  std::vector<int> dilations = ctx->Attrs().Get<std::vector<int>>("dilations");
C
chengduoZH 已提交
45

C
chengduoZH 已提交
46 47
  PADDLE_ENFORCE(in_dims.size() == 4 || in_dims.size() == 5,
                 "Conv intput should be 4-D or 5-D tensor.");
C
chengduoZH 已提交
48 49 50 51 52 53 54 55 56
  PADDLE_ENFORCE_EQ(
      in_dims.size(), filter_dims.size(),
      "Conv input dimension and filter dimension should be the same.");
  PADDLE_ENFORCE(
      in_dims.size() - strides.size() == 2U,
      "Conv input dimension and strides dimension should be consistent.");
  PADDLE_ENFORCE_EQ(
      paddings.size(), strides.size(),
      "Conv paddings dimension and Conv strides dimension should be the same.");
F
fengjiayi 已提交
57

Y
Yang Yu 已提交
58
  PADDLE_ENFORCE_EQ(in_dims[1], filter_dims[1] * groups,
C
chengduoZH 已提交
59
                    "The number of input channels should be equal to filter "
C
chengduoZH 已提交
60
                    "channels * groups.");
C
chengduoZH 已提交
61
  PADDLE_ENFORCE_EQ(
Y
Yang Yu 已提交
62
      filter_dims[0] % groups, 0,
C
chengduoZH 已提交
63 64 65
      "The number of output channels should be divided by groups.");

  std::vector<int64_t> output_shape({in_dims[0], filter_dims[0]});
C
chengduoZH 已提交
66
  for (size_t i = 0; i < strides.size(); ++i) {
Y
Yang Yang 已提交
67 68 69
    output_shape.push_back(ConvOutputSize(in_dims[i + 2], filter_dims[i + 2],
                                          dilations[i], paddings[i],
                                          strides[i]));
C
chengduoZH 已提交
70
  }
71
  ctx->SetOutputDim("Output", framework::make_ddim(output_shape));
Y
Yang Yu 已提交
72
  ctx->ShareLoD("Input", "Output");
C
chengduoZH 已提交
73 74
}

75 76
framework::OpKernelType ConvOp::GetExpectedKernelType(
    const framework::ExecutionContext& ctx) const {
77
  framework::LibraryType library{framework::LibraryType::kPlain};
M
mozga-intel 已提交
78
  // TODO(pzelazko-intel): enable MKLDNN layout when it's ready
79
  std::string data_format = ctx.Attr<std::string>("data_format");
M
mozga-intel 已提交
80 81
  framework::DataLayout layout = framework::StringToDataLayout(data_format);

C
chengduoZH 已提交
82
#ifdef PADDLE_WITH_CUDA
83
  if (platform::CanCUDNNBeUsed(ctx)) {
84
    library = framework::LibraryType::kCUDNN;
C
chengduoZH 已提交
85 86
  }
#endif
87
#ifdef PADDLE_WITH_MKLDNN
88
  if (library == framework::LibraryType::kPlain &&
89
      platform::CanMKLDNNBeUsed(ctx)) {
90
    library = framework::LibraryType::kMKLDNN;
M
mozga-intel 已提交
91
    layout = framework::DataLayout::kMKLDNN;
92
  }
93
#endif
94

K
Kexin Zhao 已提交
95 96 97 98 99 100 101 102
  auto input_data_type =
      framework::ToDataType(ctx.Input<Tensor>("Input")->type());
  auto filter_data_type =
      framework::ToDataType(ctx.Input<Tensor>("Filter")->type());
  PADDLE_ENFORCE_EQ(input_data_type, filter_data_type,
                    "input and filter data type should be consistent");

  if (input_data_type == framework::proto::VarType::FP16) {
103
    PADDLE_ENFORCE_EQ(library, framework::LibraryType::kCUDNN,
K
Kexin Zhao 已提交
104 105 106
                      "float16 can only be used when CUDNN is used");
  }

107 108
  return framework::OpKernelType(input_data_type, ctx.GetPlace(), layout,
                                 library);
109 110
}

Y
Yu Yang 已提交
111
void Conv2DOpMaker::Make() {
K
Krzysztof Binias 已提交
112
  AddAttr<bool>("is_test", "").SetDefault(false);
C
chengduoZH 已提交
113 114
  AddInput(
      "Input",
C
fix doc  
chengduoZH 已提交
115 116 117 118
      "(Tensor) The input tensor of convolution operator. "
      "The format of input tensor is NCHW, where N is batch size, C is the "
      "number of channels, H is the height of the feature, "
      "and W is the width of the feature.");
C
chengduoZH 已提交
119
  AddInput("Filter",
C
fix doc  
chengduoZH 已提交
120
           "(Tensor) The filter tensor of convolution operator. "
C
chengduoZH 已提交
121 122
           "The format of the filter tensor is MCHW, where M is the number of "
           "output image channels, C is the number of input image channels, "
C
fix doc  
chengduoZH 已提交
123 124
           "H is the height of the filter, and W is the width of the filter. "
           "If the groups attribute is greater than 1, C equals the number of "
C
chengduoZH 已提交
125
           "input image channels divided by the groups.");
126 127 128 129 130
  AddInput("Bias",
           "(Tensor) Bias to be added to each output of filter application."
           "The format of output tensor is X (one-dimensional) of size equal"
           "to the number of output channels. Only used with MKL-DNN.")
      .AsDispensable();
C
chengduoZH 已提交
131
  AddOutput("Output",
C
fix doc  
chengduoZH 已提交
132
            "(Tensor) The output tensor of convolution operator. "
133 134
            "The format of output tensor is also NCHW.")
      .Reuse("Input");
135 136 137
  AddInput("ResidualData",
           "(Tensor) Tensor with residual data "
           "to which convolution output will be added."
138
           "Used with fuse_residual_connection fusion.")
139
      .AsDispensable();
C
chengduoZH 已提交
140 141 142 143
  AddAttr<std::vector<int>>("strides",
                            "(vector<int> default:{1, 1}), the "
                            "strides(h_stride, w_stride) of "
                            "convolution operator.")
C
chengduoZH 已提交
144
      .SetDefault({1, 1});
C
chengduoZH 已提交
145 146 147 148
  AddAttr<std::vector<int>>("paddings",
                            "(vector<int> default:{0, 0}), the "
                            "paddings(h_pad, w_pad) of "
                            "convolution operator.")
C
chengduoZH 已提交
149 150 151
      .SetDefault({0, 0});
  AddAttr<int>(
      "groups",
C
chengduoZH 已提交
152
      "(int default:1), the groups number of the convolution operator. "
C
fix doc  
chengduoZH 已提交
153 154 155 156
      "According to grouped convolution in Alex Krizhevsky's Deep CNN paper: "
      "when group=2, the first half of the filters is only connected to the "
      "first half of the input channels, while the second half of the filters "
      "is only connected to the second half of the input channels.")
C
chengduoZH 已提交
157
      .SetDefault(1);
C
chengduoZH 已提交
158
  AddAttr<std::vector<int>>("dilations",
C
chengduoZH 已提交
159 160
                            "(vector<int> default:{1, 1}), the "
                            "dilations(h_dilation, w_dilation) of "
C
chengduoZH 已提交
161
                            "convolution operator.")
C
chengduoZH 已提交
162
      .SetDefault({1, 1});
163 164 165 166
  AddAttr<bool>(
      "use_cudnn",
      "(bool, default false) Only used in cudnn kernel, need install cudnn")
      .SetDefault(false);
167 168 169
  AddAttr<bool>("use_mkldnn",
                "(bool, default false) Only used in mkldnn kernel")
      .SetDefault(false);
M
Michal Gallus 已提交
170 171
  AddAttr<bool>("fuse_relu", "(bool, default false) Only used in mkldnn kernel")
      .SetDefault(false);
172
  AddAttr<bool>("fuse_residual_connection",
173
                "(bool, default false) Only used in mkldnn kernel. Used "
174 175
                "whenever convolution output is as an input to residual "
                "connection.")
176
      .SetDefault(false);
177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192
  AddAttr<std::string>(
      "data_format",
      "(string, default NCHW) Only used in "
      "An optional string from: \"NHWC\", \"NCHW\". "
      "Defaults to \"NHWC\". Specify the data format of the output data, "
      "the input will be transformed automatically. ")
      .SetDefault("AnyLayout");
  // TODO(dzhwinter): need to registered layout transform function
  AddAttr<int>("workspace_size_MB",
               "Only used in cudnn kernel. Need set use_cudnn to true."
               "workspace size for cudnn, in MB, "
               "workspace is a section of GPU memory which will be "
               "allocated/freed each time the operator runs, larger "
               "workspace size can increase performance but also requires "
               "better hardware. This size should be chosen carefully.")
      .SetDefault(4096);
C
chengduoZH 已提交
193
  AddComment(R"DOC(
C
fix doc  
chengduoZH 已提交
194 195
Convolution Operator.

C
chengduoZH 已提交
196
The convolution operation calculates the output based on the input, filter
C
chengduoZH 已提交
197
and strides, paddings, dilations, groups parameters. The size of each dimension of the
C
chengduoZH 已提交
198
parameters is checked in the infer-shape.
C
chengduoZH 已提交
199
Input(Input) and Output(Output) are in NCHW format. Where N is batch
C
fix doc  
chengduoZH 已提交
200
size, C is the number of channels, H is the height of the feature, and W is
C
chengduoZH 已提交
201 202 203 204 205 206
the width of the feature.
Filters(Input) is MCHW format. Where M is the number of output image channels, C is
the number of input image channels, H is the height of the filter, and W
is the width of the filter.
Parameters(strides, paddings, dilations) are two elements. These two elements represent
height and width, respectively.
C
chengduoZH 已提交
207 208 209 210
The input(X) size and output(Out) size may be different.

Example:
  Input:
C
chengduoZH 已提交
211 212
       Input shape: $(N, C_{in}, H_{in}, W_{in})$
       Filter shape: $(C_{out}, C_{in}, H_f, W_f)$
C
chengduoZH 已提交
213
  Output:
C
chengduoZH 已提交
214 215 216 217 218 219
       Output shape: $(N, C_{out}, H_{out}, W_{out})$
  Where
$$
       H_{out}= \frac{(H_{in} + 2 * paddings[0] - (dilations[0] * (H_f - 1) + 1))}{strides[0]}+ 1 \\
       W_{out}= \frac{(W_{in} + 2 * paddings[1] - (dilations[1] * (W_f - 1) + 1))}{strides[1]}+ 1
$$
C
chengduoZH 已提交
220
)DOC");
C
chengduoZH 已提交
221 222
}

Y
Yu Yang 已提交
223
void Conv3DOpMaker::Make() {
C
chengduoZH 已提交
224 225
  AddInput(
      "Input",
C
fix doc  
chengduoZH 已提交
226
      "(Tensor) The input tensor of convolution operator. "
C
chengduoZH 已提交
227
      "The format of input tensor is NCDHW. Where N is batch size, C is the "
C
fix doc  
chengduoZH 已提交
228 229 230
      "number of channels, D is the depth of the feature, H is the height of "
      "the feature, "
      "and W is the width of the feature.");
C
chengduoZH 已提交
231
  AddInput("Filter",
C
fix doc  
chengduoZH 已提交
232
           "(Tensor) The filter tensor of convolution operator. "
C
chengduoZH 已提交
233 234
           "The format of the filter tensor is MCDHW, where M is the number of "
           "output image channels, C is the number of input image channels, "
C
fix doc  
chengduoZH 已提交
235 236 237
           "D is the depth of the filter, H is the height of the filter, and W "
           "is the width of the filter."
           "If the groups attribute is greater than 1, C equals the number of "
C
chengduoZH 已提交
238 239
           "input image channels divided by the groups.");
  AddOutput("Output",
C
fix doc  
chengduoZH 已提交
240
            "(Tensor) The output tensor of convolution operator."
241 242
            "The format of output tensor is also NCDHW.")
      .Reuse("Input");
C
chengduoZH 已提交
243 244 245 246
  AddAttr<std::vector<int>>("strides",
                            "(vector<int>, default:{1, 1, 1}), the "
                            "strides(d_stride, h_stride, w_stride) of "
                            "convolution operator.")
C
chengduoZH 已提交
247
      .SetDefault({1, 1, 1});
C
chengduoZH 已提交
248 249 250 251
  AddAttr<std::vector<int>>("paddings",
                            "(vector<int>, default:{0, 0, 0}), the "
                            "paddings(d_pad, h_pad, w_pad) of convolution "
                            "operator.")
C
chengduoZH 已提交
252 253 254
      .SetDefault({0, 0, 0});
  AddAttr<int>(
      "groups",
C
chengduoZH 已提交
255
      "(int default:1), the groups number of the convolution operator. "
C
fix doc  
chengduoZH 已提交
256 257 258 259
      "According to grouped convolution in Alex Krizhevsky's Deep CNN paper: "
      "when group=2, the first half of the filters is only connected to the "
      "first half of the input channels, while the second half of the filters "
      "is only connected to the second half of the input channels.")
C
chengduoZH 已提交
260
      .SetDefault(1);
C
chengduoZH 已提交
261
  AddAttr<std::vector<int>>("dilations",
C
chengduoZH 已提交
262 263
                            "(vector<int> default:{1, 1, 1}), the "
                            "dilations(d_dilation, h_dilation, w_dilation) of "
C
chengduoZH 已提交
264
                            "convolution operator.")
C
chengduoZH 已提交
265
      .SetDefault({1, 1, 1});
266 267 268 269
  AddAttr<bool>(
      "use_cudnn",
      "(bool, default false) Only used in cudnn kernel, need install cudnn")
      .SetDefault(false);
270 271 272
  AddAttr<bool>("use_mkldnn",
                "(bool, default false) Only used in mkldnn kernel")
      .SetDefault(false);
273 274 275 276 277 278 279 280 281 282 283 284 285 286 287
  AddAttr<std::string>(
      "data_format",
      "(string, default NCHW) Only used in "
      "An optional string from: \"NHWC\", \"NCHW\". "
      "Defaults to \"NHWC\". Specify the data format of the output data, "
      "the input will be transformed automatically. ")
      .SetDefault("AnyLayout");
  // TODO(dzhwinter): need to registered layout transform function
  AddAttr<int>("workspace_size_MB",
               "Only used in cudnn kernel. workspace size for cudnn, in MB, "
               "workspace is a section of GPU memory which will be "
               "allocated/freed each time the operator runs, larger "
               "workspace size can increase performance but also requires "
               "better hardware. This size should be chosen carefully.")
      .SetDefault(4096);
C
fix doc  
chengduoZH 已提交
288

C
chengduoZH 已提交
289
  AddComment(R"DOC(
C
fix doc  
chengduoZH 已提交
290 291
Convolution3D Operator.

C
chengduoZH 已提交
292
The convolution operation calculates the output based on the input, filter
C
chengduoZH 已提交
293
and strides, paddings, dilations, groups parameters. The size of each dimension of the
C
chengduoZH 已提交
294
parameters is checked in the infer-shape.
C
chengduoZH 已提交
295
Input(Input) and output(Output) are in NCDHW format, where N is batch
C
fix doc  
chengduoZH 已提交
296
size, C is the number of channels,D is the depth of the feature, H is the height of
C
chengduoZH 已提交
297 298 299 300 301 302
the feature, and W is the width of the feature.
Filters(Input) is MCDHW format, where M is the number of output image channels,
C is the number of input image channels, D is the depth of the filter,
H is the height of the filter, and W is the width of the filter.
Parameters(strides, paddings, dilations) are three elements. These three elements
represent depth, height and width, respectively.
C
fix doc  
chengduoZH 已提交
303 304 305 306
The input(X) size and output(Out) size may be different.

Example:
  Input:
C
chengduoZH 已提交
307 308
       Input shape: $(N, C_{in}, D_{in}, H_{in}, W_{in})$
       Filter shape: $(C_{out}, C_{in}, D_f, H_f, W_f)$
C
fix doc  
chengduoZH 已提交
309
  Output:
C
chengduoZH 已提交
310 311 312 313 314 315 316
       Output shape: $(N, C_{out}, D_{out}, H_{out}, W_{out})$
  Where
  $$
       D_{out}= \frac{(D_{in} + 2 * paddings[0] - (dilations[0] * (D_f - 1) + 1))}{ strides[0]}+ 1 \\
       H_{out}= \frac{(H_{in} + 2 * paddings[1] - (dilations[1] * (H_f - 1) + 1))}{ strides[1]}+ 1 \\
       W_{out}= \frac{(W_{in} + 2 * paddings[2] - (dilations[2] * (W_f - 1) + 1))}{ strides[2]}+ 1
  $$
C
chengduoZH 已提交
317 318 319
)DOC");
}

C
chengduoZH 已提交
320 321 322 323 324 325 326 327 328 329 330
void ConvOpGrad::InferShape(framework::InferShapeContext* ctx) const {
  auto in_dims = ctx->GetInputDim("Input");
  auto filter_dims = ctx->GetInputDim("Filter");
  if (ctx->HasOutput(framework::GradVarName("Input"))) {
    ctx->SetOutputDim(framework::GradVarName("Input"), in_dims);
  }
  if (ctx->HasOutput(framework::GradVarName("Filter"))) {
    ctx->SetOutputDim(framework::GradVarName("Filter"), filter_dims);
  }
}

331 332
framework::OpKernelType ConvOpGrad::GetExpectedKernelType(
    const framework::ExecutionContext& ctx) const {
333
  framework::LibraryType library_{framework::LibraryType::kPlain};
M
mozga-intel 已提交
334 335 336 337
  // TODO(pzelazko-intel): enable MKLDNN layout when it's ready
  std::string data_format = ctx.Attr<std::string>("data_format");
  framework::DataLayout layout_ = framework::StringToDataLayout(data_format);

C
chengduoZH 已提交
338
#ifdef PADDLE_WITH_CUDA
339 340
  if (platform::CanCUDNNBeUsed(ctx)) {
    library_ = framework::LibraryType::kCUDNN;
C
chengduoZH 已提交
341 342
  }
#endif
343 344 345 346
#ifdef PADDLE_WITH_MKLDNN
  if (library_ == framework::LibraryType::kPlain &&
      platform::CanMKLDNNBeUsed(ctx)) {
    library_ = framework::LibraryType::kMKLDNN;
M
mozga-intel 已提交
347
    layout_ = framework::DataLayout::kMKLDNN;
348
  }
349
#endif
350 351 352 353 354 355

  return framework::OpKernelType(
      framework::ToDataType(ctx.Input<Tensor>("Input")->type()), ctx.GetPlace(),
      layout_, library_);
}

C
chengduoZH 已提交
356 357 358 359
}  // namespace operators
}  // namespace paddle

namespace ops = paddle::operators;
Y
Yang Yang 已提交
360
REGISTER_OPERATOR(conv2d, ops::ConvOp, ops::Conv2DOpMaker,
361 362
                  paddle::framework::DefaultGradOpDescMaker<true>);
REGISTER_OPERATOR(conv2d_grad, ops::ConvOpGrad);
363 364

// depthwise convolution op
Y
Yang Yang 已提交
365
REGISTER_OPERATOR(depthwise_conv2d, ops::ConvOp, ops::Conv2DOpMaker,
366 367
                  paddle::framework::DefaultGradOpDescMaker<true>);
REGISTER_OPERATOR(depthwise_conv2d_grad, ops::ConvOpGrad);
Y
Yang Yang 已提交
368
REGISTER_OPERATOR(conv3d, ops::ConvOp, ops::Conv3DOpMaker,
369 370
                  paddle::framework::DefaultGradOpDescMaker<true>);
REGISTER_OPERATOR(conv3d_grad, ops::ConvOpGrad);
C
chengduoZH 已提交
371

372 373
// depthwise conv kernel
// TODO(xingzhaolong): neon kernel for mobile
Z
zlx 已提交
374
REGISTER_OP_CPU_KERNEL(
375
    depthwise_conv2d,
X
xzl 已提交
376 377 378 379
    ops::GemmConvKernel<paddle::platform::CPUDeviceContext, float>,
    ops::GemmConvKernel<paddle::platform::CPUDeviceContext, double>);

REGISTER_OP_CPU_KERNEL(
380
    depthwise_conv2d_grad,
X
xzl 已提交
381 382
    ops::GemmConvGradKernel<paddle::platform::CPUDeviceContext, float>,
    ops::GemmConvGradKernel<paddle::platform::CPUDeviceContext, double>);
Z
zlx 已提交
383

C
chengduoZH 已提交
384
REGISTER_OP_CPU_KERNEL(
Q
QI JUN 已提交
385 386 387 388 389 390
    conv2d, ops::GemmConvKernel<paddle::platform::CPUDeviceContext, float>,
    ops::GemmConvKernel<paddle::platform::CPUDeviceContext, double>);
REGISTER_OP_CPU_KERNEL(
    conv2d_grad,
    ops::GemmConvGradKernel<paddle::platform::CPUDeviceContext, float>,
    ops::GemmConvGradKernel<paddle::platform::CPUDeviceContext, double>);
C
chengduoZH 已提交
391 392

REGISTER_OP_CPU_KERNEL(
Q
QI JUN 已提交
393 394 395 396 397 398
    conv3d, ops::GemmConvKernel<paddle::platform::CPUDeviceContext, float>,
    ops::GemmConvKernel<paddle::platform::CPUDeviceContext, double>);
REGISTER_OP_CPU_KERNEL(
    conv3d_grad,
    ops::GemmConvGradKernel<paddle::platform::CPUDeviceContext, float>,
    ops::GemmConvGradKernel<paddle::platform::CPUDeviceContext, double>);