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
|
#include "conv_relu_op.h"
namespace caffe2 {
template <typename T, class Context>
bool ConvReluOp<T, Context>::RunOnDeviceWithOrderNCHW() {
// Delegate to local conv operator
for (int i = 0; i < this->InputSize(); ++i) {
local_input_blobs_[i]->ShareExternal(
// NOLINTNEXTLINE(cppcoreguidelines-pro-type-const-cast)
const_cast<void*>(this->Inputs()[i]->GetRaw()),
this->Inputs()[i]->meta());
}
if (!local_op_->RunOnDeviceWithOrderNCHW()) {
return false;
}
// Apply Relu
Tensor* local_output =
BlobGetMutableTensor(local_output_blobs_[0], Context::GetDeviceType());
const T* output_local_data = local_output->template data<T>();
Tensor* output =
Operator<Context>::Output(0, local_output->sizes(), at::dtype<T>());
T* output_data = output->template mutable_data<T>();
#ifdef _OPENMP
#pragma omp parallel for
#endif
for (int i = 0; i < output->numel(); ++i) {
output_data[i] = std::max(static_cast<T>(0), output_local_data[i]);
}
return true;
}
template <typename T, class Context>
bool ConvReluOp<T, Context>::RunOnDeviceWithOrderNHWC() {
// Delegate to local conv operator
for (int i = 0; i < this->InputSize(); ++i) {
local_input_blobs_[i]->ShareExternal(
// NOLINTNEXTLINE(cppcoreguidelines-pro-type-const-cast)
const_cast<void*>(this->Inputs()[i]->GetRaw()),
this->Inputs()[i]->meta());
}
if (!local_op_->RunOnDeviceWithOrderNHWC()) {
return false;
}
// Apply Relu
Tensor* local_output =
BlobGetMutableTensor(local_output_blobs_[0], Context::GetDeviceType());
const T* output_local_data = local_output->template data<T>();
Tensor* output =
Operator<Context>::Output(0, local_output->sizes(), at::dtype<T>());
T* output_data = output->template mutable_data<T>();
#ifdef _OPENMP
#pragma omp parallel for
#endif
for (int i = 0; i < output->numel(); ++i) {
output_data[i] = std::max(static_cast<T>(0), output_local_data[i]);
}
return true;
}
OPERATOR_SCHEMA(ConvRelu)
.NumInputs(2, 3)
.NumOutputs(1)
.TensorInferenceFunction(ConvPoolOpBase<CPUContext>::TensorInferenceForConv)
.CostInferenceFunction(OpSchema::CostInferenceFunctionType(
ConvPoolOpBase<CPUContext>::CostInferenceForConv));
REGISTER_CPU_OPERATOR(ConvRelu, ConvReluOp<float, CPUContext>);
} // namespace caffe2
|
