fusion_conv_bn_relu_op.h 3.2 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 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
/* Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved.

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. */

#ifdef FUSION_CONVBNRELU_OP

#pragma once

#include <string>
#include <vector>
#include "framework/operator.h"
#include "framework/program/program-optimize/fusion_op_register.h"
#include "operators/kernel/conv_bn_relu_kernel.h"
#include "operators/op_param.h"

namespace paddle_mobile {
namespace operators {
using std::string;
using std::vector;
class FusionConvBNReluMatcher : public framework::FusionOpMatcher {
 public:
  FusionConvBNReluMatcher() {
    node_ = framework::Node(G_OP_TYPE_CONV);
    node_ > std::make_shared<framework::Node>(G_OP_TYPE_BATCHNORM) >
        std::make_shared<framework::Node>(G_OP_TYPE_RELU);
  }

  void FolderNodes(
      framework::Node *node,
      std::vector<std::shared_ptr<framework::Node>> *removed_nodes) {
    node->Folder(node_.Depth(), Type(),
                 {{G_OP_TYPE_BATCHNORM,
                   {{"Scale", "Scale"},
                    {"Mean", "Mean"},
                    {"Bias", "Bias"},
                    {"Variance", "Variance"}}}},
                 removed_nodes);
  }

  std::string Type() { return G_OP_TYPE_FUSION_CONV_BN_RELU; }
};

template <typename DeviceType, typename T>
class FusionConvBNReluOp : public framework::OperatorWithKernel<
N
nhzlx 已提交
55
                               DeviceType, FusionConvBNReluParam<DeviceType>,
56 57 58 59 60 61 62
                               operators::ConvBNReluKernel<DeviceType, T>> {
 public:
  FusionConvBNReluOp(const string &type, const VariableNameMap &inputs,
                     const VariableNameMap &outputs,
                     const framework::AttributeMap &attrs,
                     std::shared_ptr<framework::Scope> scope)
      : framework::OperatorWithKernel<
N
nhzlx 已提交
63
            DeviceType, FusionConvBNReluParam<DeviceType>,
64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85
            operators::ConvBNReluKernel<DeviceType, T>>(type, inputs, outputs,
                                                        attrs, scope) {}
  void InferShape() const override;

 protected:
};

#ifdef PADDLE_MOBILE_CPU

#ifndef FUSION_CONV_BN_RELU_REGISTER
static framework::FusionOpRegistrar fusion_conv_bn_relu_registrar(
    new FusionConvBNReluMatcher());
#define FUSION_CONV_BN_RELU_REGISTER
#endif

#endif

#ifdef PADDLE_MOBILE_MALI_GPU

#endif

#ifdef PADDLE_MOBILE_FPGA
Z
zhangyang 已提交
86 87 88 89 90 91

#ifndef FUSION_CONV_BN_RELU_REGISTER
static framework::FusionOpRegistrar fusion_conv_bn_relu_registrar(
    new FusionConvBNReluMatcher());
#define FUSION_CONV_BN_RELU_REGISTER
#endif
92 93 94 95 96 97 98 99 100 101 102
#endif

}  // namespace operators
}  // namespace paddle_mobile

#ifdef PADDLE_MOBILE_CPU
USE_OP_CPU(fusion_conv_bn_relu);
#endif
#ifdef PADDLE_MOBILE_MALI_GPU
#endif
#ifdef PADDLE_MOBILE_FPGA
Z
zhangyang 已提交
103
USE_OP_FPGA(fusion_conv_bn_relu);
104 105 106
#endif

#endif