conv_transpose_cudnn_op.cc 3.1 KB
Newer Older
Z
zchen0211 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14
/* 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. */

C
chengduoZH 已提交
15
#include "paddle/operators/conv_transpose_op.h"
Z
zchen0211 已提交
16 17 18 19 20 21 22 23 24 25

namespace paddle {
namespace operators {

class CudnnConv2DTransposeOpMaker : public Conv2DTransposeOpMaker {
 public:
  CudnnConv2DTransposeOpMaker(framework::OpProto* proto,
                              framework::OpAttrChecker* op_checker)
      : Conv2DTransposeOpMaker(proto, op_checker) {
    AddAttr<std::vector<int>>("dilations", "dilations of convolution operator.")
C
chengduoZH 已提交
26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43
        .SetDefault({1, 1});
    AddAttr<int>("workspace_size_MB",
                 "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 hardward. This size should be carefully setted.")
        .SetDefault(4096);
  }
};

class CudnnConv3DTransposeOpMaker : public Conv3DTransposeOpMaker {
 public:
  CudnnConv3DTransposeOpMaker(framework::OpProto* proto,
                              framework::OpAttrChecker* op_checker)
      : Conv3DTransposeOpMaker(proto, op_checker) {
    AddAttr<std::vector<int>>("dilations", "dilations of convolution operator.")
        .SetDefault({1, 1, 1});
Z
zchen0211 已提交
44 45 46 47 48 49 50 51 52 53 54 55 56 57
    AddAttr<int>("workspace_size_MB",
                 "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 hardward. This size should be carefully setted.")
        .SetDefault(4096);
  }
};

}  // namespace operators
}  // namespace paddle

namespace ops = paddle::operators;
C
chengduoZH 已提交
58
REGISTER_OP(conv2d_transpose_cudnn, ops::ConvTransposeOp,
Z
zchen0211 已提交
59
            ops::CudnnConv2DTransposeOpMaker, conv2d_transpose_cudnn_grad,
C
chengduoZH 已提交
60
            ops::ConvTransposeOpGrad);
Z
zchen0211 已提交
61 62

REGISTER_OP_CPU_KERNEL(
Z
zchen0211 已提交
63
    conv2d_transpose_cudnn,
64
    ops::GemmConvTransposeKernel<paddle::platform::CPUPlace, float>);
Z
zchen0211 已提交
65
REGISTER_OP_CPU_KERNEL(
Z
zchen0211 已提交
66
    conv2d_transpose_cudnn_grad,
67
    ops::GemmConvTransposeGradKernel<paddle::platform::CPUPlace, float>);
C
chengduoZH 已提交
68 69 70 71 72 73 74 75 76 77 78

REGISTER_OP(conv3d_transpose_cudnn, ops::ConvTransposeOp,
            ops::CudnnConv3DTransposeOpMaker, conv3d_transpose_cudnn_grad,
            ops::ConvTransposeOpGrad);

REGISTER_OP_CPU_KERNEL(
    conv3d_transpose_cudnn,
    ops::GemmConvTransposeKernel<paddle::platform::CPUPlace, float>);
REGISTER_OP_CPU_KERNEL(
    conv3d_transpose_cudnn_grad,
    ops::GemmConvTransposeGradKernel<paddle::platform::CPUPlace, float>);