File: elu_op.cc

package info (click to toggle)
pytorch 1.13.1%2Bdfsg-4
  • links: PTS, VCS
  • area: main
  • in suites: bookworm
  • size: 139,252 kB
  • sloc: cpp: 1,100,274; python: 706,454; ansic: 83,052; asm: 7,618; java: 3,273; sh: 2,841; javascript: 612; makefile: 323; xml: 269; ruby: 185; yacc: 144; objc: 68; lex: 44
file content (146 lines) | stat: -rw-r--r-- 3,865 bytes parent folder | download
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
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
#include "caffe2/operators/elu_op.h"

#include <algorithm>
#include <functional>
#include <string>

#include "caffe2/utils/eigen_utils.h"

namespace caffe2 {

template <>
template <typename T>
bool EluFunctor<CPUContext>::
operator()(const int N, const T* X, T* Y, CPUContext* /* context */) const {
  ConstEigenVectorArrayMap<T> X_arr(X, N);
  EigenVectorMap<T>(Y, N) =
      (X_arr < 0).select(alpha * (X_arr.exp() - T(1)), X_arr);
  return true;
}

template <>
template <typename T>
bool EluGradientFunctor<CPUContext>::Forward(
    const std::vector<int>& Y_dims,
    const std::vector<int>& /* dY_dims */,
    const T* Y,
    const T* dY,
    T* dX,
    CPUContext* /* context */) const {
  const int size = std::accumulate(
      // NOLINTNEXTLINE(modernize-use-transparent-functors)
      Y_dims.cbegin(), Y_dims.cend(), 1, std::multiplies<int>());
  ConstEigenVectorArrayMap<T> Y_arr(Y, size);
  ConstEigenVectorArrayMap<T> dY_arr(dY, size);
  EigenVectorArrayMap<T>(dX, size) =
      (Y_arr < 0).select(dY_arr * (Y_arr + alpha), dY_arr);
  return true;
}

REGISTER_CPU_OPERATOR(
    Elu,
    UnaryElementwiseWithArgsOp<
        TensorTypes<float>,
        CPUContext,
        EluFunctor<CPUContext>>);
REGISTER_CPU_GRADIENT_OPERATOR(
    EluGradient,
    BinaryElementwiseWithArgsOp<
        TensorTypes<float>,
        CPUContext,
        EluGradientFunctor<CPUContext>>);

// Input: X, output: Y
OPERATOR_SCHEMA(Elu)
    .NumInputs(1)
    .NumOutputs(1)
    .AllowInplace({{0, 0}})
    .IdenticalTypeAndShape()
    .SetDoc(R"DOC(

This op implements the exponential linear unit (ELU) activation function as described in [Fast and Accurate Deep Network Learning by Exponential Linear Units (ELUs)](https://arxiv.org/abs/1511.07289). The op takes an input tensor $X$ of arbitrary shape, computes the elementwise elu operation, and returns a vector $Y$ of the same shape as output. The alpha parameter may be passed as an argument, but defaults to 1. The elu operation is defined as

$$y=f(x) =\begin{cases}\alpha(e^x-1) & x < 0 \\ x & otherwise\end{cases}$$

Github Links:
- https://github.com/pytorch/pytorch/blob/master/caffe2/operators/elu_op.h
- https://github.com/pytorch/pytorch/blob/master/caffe2/operators/elu_op.cc

<details>

<summary> <b>Example</b> </summary>

**Code**

```
workspace.ResetWorkspace()

op = core.CreateOperator(
    "Elu",
    ["X"],
    ["Y"],
    alpha=1.1
)

workspace.FeedBlob("X", np.random.randn(3, 3).astype(np.float32))
print("X:\n", workspace.FetchBlob("X"), "\n")

workspace.RunOperatorOnce(op)
print("Y:\n", workspace.FetchBlob("Y"))

```

**Result**

```

X:
 [[ 0.35339102  1.1860217  -0.10710736]
 [-3.1173866  -0.1889988  -0.20330353]
 [ 1.8525308  -0.368949    0.506277  ]]

Y:
 [[ 0.35339102  1.1860217  -0.11172786]
 [-1.0513     -0.18943374 -0.20236646]
 [ 1.8525308  -0.33939326  0.506277  ]]

```

</details>

)DOC")
    .Input(0, "X", "1D input tensor of data to be operated on.")
    .Output(0, "Y", "1D input tensor, calculated as described above.")
    .Arg(
        "alpha",
        "*(type: float; default: 1.0)* Defines alpha parameter used in calculation.")
    .InheritOnnxSchema();

// Input: Y, dY, output: dX
GRADIENT_OPERATOR_SCHEMA(EluGradient)
    .NumInputs(2)
    .NumOutputs(1)
    .AllowInplace({{1, 0}})
    .SetDoc(R"DOC(
EluGradient takes both Y and dY and uses this to update dX according to the
chain rule and derivatives of the rectified linear function.
)DOC");

namespace {

class GetEluGradient : public GradientMakerBase {
  using GradientMakerBase::GradientMakerBase;
  std::vector<OperatorDef> GetGradientDefs() override {
    return SingleGradientDef(
        def_.type() + "Gradient",
        "",
        std::vector<std::string>{O(0), GO(0)},
        std::vector<std::string>{GI(0)});
  }
};

} // namespace

REGISTER_GRADIENT(Elu, GetEluGradient);

} // namespace caffe2