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
|
/*
* Copyright (c) 2018 Samsung Electronics Co., Ltd. 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.
*/
#include <OperationsUtils.h>
#include <NeuralNetworks.h>
#include <arm_compute/core/TensorShape.h>
#include <arm_compute/core/TensorInfo.h>
#include "../IO_accessor.h"
#include "../shape.h"
#include "../util.h"
#include "../NEUniqueTensor.h"
namespace nnfw {
namespace kernel {
namespace acl {
namespace neon {
bool softmaxFloat32(const float* inputData, const nnfw::rt::Shape& inputShape,
const float beta,
float* outputData, const nnfw::rt::Shape& outputShape)
{
arm_compute::TensorShape input_shape = util::fromNNShape(inputShape);
arm_compute::TensorShape output_shape = util::fromNNShape(outputShape);
NEUniqueTensor input(arm_compute::TensorInfo(input_shape, arm_compute::Format::F32));
NEUniqueTensor output(arm_compute::TensorInfo(output_shape, arm_compute::Format::F32));
auto softmax_f = std::make_shared<arm_compute::NESoftmaxLayer>();
softmax_f->configure(input.ptr(), output.ptr(), beta);
input.allocate();
output.allocate();
if (inputShape.dimensions.size() == 4)
{
TensorAccess<InputAccessor>(input.ref(), inputData, inputShape);
softmax_f->run();
TensorAccess<OutputAccessor>(output.ref(), outputData, outputShape);
}
else if (inputShape.dimensions.size() == 2)
{
// Softmax comes with 1xN matrix and this is translated to N vector in arm_compute::TensorShape
TensorAccess<VectorInputAccessor>(input.ref(), inputData, inputShape);
softmax_f->run();
TensorAccess<VectorOutputAccessor>(output.ref(), outputData, outputShape);
}
else
{
assert("undefined dimension of input" && 0);
return false;
}
return true;
}
} // namespace neon
} // namespace acl
} // namespace kernel
} // namespace nnfw
|
