elementwise_op.h 5.8 KB
Newer Older
1
/* Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserved.
G
gongweibao 已提交
2

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
G
gongweibao 已提交
6

7
    http://www.apache.org/licenses/LICENSE-2.0
G
gongweibao 已提交
8

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. */
G
gongweibao 已提交
14 15

#pragma once
16
#include <string>
Y
Yi Wang 已提交
17 18
#include "paddle/fluid/framework/op_registry.h"
#include "paddle/fluid/framework/operator.h"
G
gongweibao 已提交
19 20 21 22 23 24 25 26 27

namespace paddle {
namespace operators {

class ElementwiseOp : public framework::OperatorWithKernel {
 public:
  using framework::OperatorWithKernel::OperatorWithKernel;

  using Tensor = framework::Tensor;
28
  void InferShape(framework::InferShapeContext* ctx) const override {
Q
Qiao Longfei 已提交
29
    PADDLE_ENFORCE(ctx->HasInput("X"),
C
caoying03 已提交
30
                   "Input(X) of elementwise op should not be null.");
Q
Qiao Longfei 已提交
31
    PADDLE_ENFORCE(ctx->HasInput("Y"),
C
caoying03 已提交
32
                   "Input(Y) of elementwise op should not be null.");
Q
Qiao Longfei 已提交
33 34 35 36 37
    PADDLE_ENFORCE(ctx->HasOutput("Out"),
                   "Output(Out) of elementwise op should not be null.");

    auto x_dim = ctx->GetInputDim("X");
    auto y_dim = ctx->GetInputDim("Y");
G
gongweibao 已提交
38
    PADDLE_ENFORCE_GE(x_dim.size(), y_dim.size(),
39
                      "Rank of first input must >= rank of second input.");
Q
Qiao Longfei 已提交
40 41
    ctx->SetOutputDim("Out", x_dim);
    ctx->ShareLoD("X", /*->*/ "Out");
G
gongweibao 已提交
42 43 44
  }
};

45 46 47 48 49 50 51 52 53 54 55 56
class ElementwiseOpInferVarType : public framework::VarTypeInference {
 public:
  void operator()(const framework::OpDesc& op_desc,
                  framework::BlockDesc* block) const override {
    auto x_name = op_desc.Input("X")[0];
    auto out_name = op_desc.Output("Out")[0];
    auto& x = block->FindRecursiveOrCreateVar(x_name);
    auto& out = block->FindRecursiveOrCreateVar(out_name);
    out.SetType(x.GetType());
  }
};

G
gongweibao 已提交
57 58
class ElementwiseOpMaker : public framework::OpProtoAndCheckerMaker {
 public:
Y
Yu Yang 已提交
59
  void Make() final {
C
caoying03 已提交
60 61
    AddInput("X", "(Tensor), The first input tensor of elementwise op.");
    AddInput("Y", "(Tensor), The second input tensor of elementwise op.");
62
    AddOutput("Out", "The output of elementwise op.").Reuse("X");
G
gongweibao 已提交
63
    AddAttr<int>("axis",
C
caoying03 已提交
64 65
                 "(int, default -1). The start dimension index "
                 "for broadcasting Y onto X.")
G
gongweibao 已提交
66 67
        .SetDefault(-1)
        .EqualGreaterThan(-1);
Y
Yu Yang 已提交
68
    AddComment(string::Sprintf(R"DOC(
L
Luo Tao 已提交
69
Limited Elementwise %s Operator
K
kexinzhao 已提交
70 71 72

The equation is:

Y
Yu Yang 已提交
73
$$%s$$
K
kexinzhao 已提交
74

L
Luo Tao 已提交
75 76
- $X$: a tensor of any dimension. 
- $Y$: a tensor whose dimensions must be less than or equal to the dimensions of $X$.
K
kexinzhao 已提交
77 78

There are two cases for this operator:
79

L
Luo Tao 已提交
80 81
1. The shape of $Y$ is the same with $X$.
2. The shape of $Y$ is a continuous subsequence of $X$.
K
kexinzhao 已提交
82 83

For case 2:
84

L
Luo Tao 已提交
85 86 87 88 89
1. Broadcast $Y$ to match the shape of $X$, where $axis$ is the start dimension index 
   for broadcasting $Y$ onto $X$. 
2. If $axis$ is -1 (default), $axis = rank(X) - rank(Y)$.
3. The trailing dimensions of size 1 for $Y$ will be ignored for the consideration of 
   subsequence, such as shape(Y) = (2, 1) => (2).
K
kexinzhao 已提交
90

L
Luo Tao 已提交
91
For example:
92

93
  .. code-block:: python
G
gongweibao 已提交
94

95 96
    shape(X) = (2, 3, 4, 5), shape(Y) = (,)
    shape(X) = (2, 3, 4, 5), shape(Y) = (5,)
L
Luo Tao 已提交
97
    shape(X) = (2, 3, 4, 5), shape(Y) = (4, 5), with axis=-1(default) or axis=2
98 99
    shape(X) = (2, 3, 4, 5), shape(Y) = (3, 4), with axis=1
    shape(X) = (2, 3, 4, 5), shape(Y) = (2), with axis=0
100
    shape(X) = (2, 3, 4, 5), shape(Y) = (2, 1), with axis=0
101

L
Luo Tao 已提交
102 103
The inputs $X$ and $Y$ can carry the different LoD information. 
But the output only shares the LoD information with the input $X$.
K
kexinzhao 已提交
104

Y
Yu Yang 已提交
105 106
)DOC",
                               GetName(), GetEquation()));
G
gongweibao 已提交
107 108 109
  }

 protected:
Y
Yu Yang 已提交
110 111
  virtual std::string GetName() const = 0;
  virtual std::string GetEquation() const = 0;
G
gongweibao 已提交
112 113 114 115 116 117 118
};

class ElementwiseOpGrad : public framework::OperatorWithKernel {
 public:
  using framework::OperatorWithKernel::OperatorWithKernel;
  using Tensor = framework::Tensor;

119
  void InferShape(framework::InferShapeContext* ctx) const override {
Q
Qiao Longfei 已提交
120 121 122 123 124 125 126 127
    PADDLE_ENFORCE(ctx->HasInput("X"), "Input(X) should not be null");
    PADDLE_ENFORCE(ctx->HasInput("Y"), "Input(Y) should not be null");
    PADDLE_ENFORCE(ctx->HasInput(framework::GradVarName("Out")),
                   "Input(Out@GRAD) should not be null");

    auto x_dims = ctx->GetInputDim("X");
    auto y_dims = ctx->GetInputDim("Y");
    auto out_dims = ctx->GetInputDim(framework::GradVarName("Out"));
G
gongweibao 已提交
128 129

    PADDLE_ENFORCE_GE(x_dims.size(), y_dims.size(),
130
                      "Rank of first input must >= rank of second input.");
G
gongweibao 已提交
131

Q
Qiao Longfei 已提交
132 133 134 135
    auto x_grad_name = framework::GradVarName("X");
    auto y_grad_name = framework::GradVarName("Y");
    if (ctx->HasOutput(x_grad_name)) {
      ctx->SetOutputDim(x_grad_name, x_dims);
G
gongweibao 已提交
136
    }
Q
Qiao Longfei 已提交
137 138
    if (ctx->HasOutput(y_grad_name)) {
      ctx->SetOutputDim(y_grad_name, y_dims);
G
gongweibao 已提交
139 140 141 142 143
    }
  }
};
}  // namespace operators
}  // namespace paddle
Y
Yu Yang 已提交
144 145 146 147 148 149 150 151 152 153

#define REGISTER_ELEMWISE_OP(op_type, op_name, equation)                \
  class __ElemwiseOp##op_type##Maker__                                  \
      : public ::paddle::operators::ElementwiseOpMaker {                \
   protected:                                                           \
    virtual std::string GetName() const { return op_name; }             \
    virtual std::string GetEquation() const { return equation; }        \
  };                                                                    \
  REGISTER_OPERATOR(op_type, ::paddle::operators::ElementwiseOp,        \
                    __ElemwiseOp##op_type##Maker__,                     \
154
                    ::paddle::operators::ElementwiseOpInferVarType,     \
Y
Yu Yang 已提交
155 156
                    ::paddle::framework::DefaultGradOpDescMaker<true>); \
  REGISTER_OPERATOR(op_type##_grad, ::paddle::operators::ElementwiseOpGrad)