forked from ROCm/pytorch
-
Notifications
You must be signed in to change notification settings - Fork 0
/
cast_op.h
55 lines (44 loc) · 1.32 KB
/
cast_op.h
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
#pragma once
#include <c10/util/irange.h>
#include "caffe2/core/context.h"
#include "caffe2/core/logging.h"
#include "caffe2/core/operator.h"
#include "caffe2/core/types.h"
#include "caffe2/utils/cast.h"
#include "caffe2/utils/conversions.h"
#include "caffe2/utils/math.h"
namespace caffe2 {
template <class Context>
class CastOp : public Operator<Context> {
public:
USE_OPERATOR_CONTEXT_FUNCTIONS;
explicit CastOp(const OperatorDef& operator_def, Workspace* ws)
: Operator<Context>(operator_def, ws) {
const ArgumentHelper helper(operator_def);
TensorProto_DataType to = cast::GetCastDataType(helper, "to");
SetBody(to);
}
bool RunOnDevice() override {
return (this->*body_)();
}
// Allow for Context-specific implementations
void SetBody(TensorProto_DataType to);
template <typename DstType>
bool DoRunWithDstType();
template <typename DstType, typename SrcType>
bool DoRunWithType() {
auto& input = Input(0);
auto* output = Output(0);
output->ResizeLike(input);
const auto* data = input.template data<SrcType>();
auto* out = output->template mutable_data<DstType>();
auto N = input.size();
for (const auto i : c10::irange(N)) {
out[i] = static_cast<DstType>(data[i]);
}
return true;
}
private:
bool (CastOp::*body_)();
};
} // namespace caffe2