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
|
//
// Copyright © 2017 Arm Ltd. All rights reserved.
// SPDX-License-Identifier: MIT
//
#pragma once
#include "ResolveType.hpp"
#include "DataLayoutIndexed.hpp"
#include "EndToEndTestImpl.hpp"
#include "armnn/INetwork.hpp"
#include "backendsCommon/test/CommonTestUtils.hpp"
#include <Permute.hpp>
#include <boost/test/unit_test.hpp>
#include <vector>
namespace
{
template<armnn::DataType ArmnnType, typename T = armnn::ResolveType<ArmnnType>>
void PermuteDataToNCHW(const std::vector<armnn::BackendId>& backends,
const armnn::DataLayout& dataLayout,
TensorInfo& tensorInfo,
std::vector<T>& data)
{
const armnn::PermutationVector NHWCToNCHW = {0, 2, 3, 1};
tensorInfo = armnnUtils::Permuted(tensorInfo, NHWCToNCHW);
std::vector<T> tmp(data.size());
armnnUtils::Permute(tensorInfo.GetShape(), NHWCToNCHW, data.data(), tmp.data(), sizeof(T));
data = tmp;
}
template<typename armnn::DataType DataType>
armnn::INetworkPtr CreateSpaceToDepthNetwork(const armnn::TensorShape& inputShape,
const armnn::TensorShape& outputShape,
const armnn::DataLayout dataLayout,
unsigned int blockSize,
const float qScale = 1.0f,
const int32_t qOffset = 0)
{
using namespace armnn;
// Builds up the structure of the network.
INetworkPtr net(INetwork::Create());
TensorInfo inputTensorInfo(inputShape, DataType, qScale, qOffset);
armnnUtils::DataLayoutIndexed dimensionIndices(dataLayout);
if (inputShape[dimensionIndices.GetHeightIndex()] % blockSize!=0
|| inputShape[dimensionIndices.GetWidthIndex()] % blockSize!=0)
{
throw InvalidArgumentException("Input shape must be divisible by block size in all spatial dimensions");
}
SpaceToDepthDescriptor spaceToDepthDesc;
spaceToDepthDesc.m_BlockSize = blockSize;
spaceToDepthDesc.m_DataLayout = dataLayout;
IConnectableLayer* SpaceToDepth = net->AddSpaceToDepthLayer(spaceToDepthDesc, "SpaceToDepth");
IConnectableLayer* input = net->AddInputLayer(0, "input");
Connect(input, SpaceToDepth, inputTensorInfo, 0, 0);
TensorInfo outputTensorInfo(outputShape, DataType, qScale, qOffset);
IConnectableLayer* output = net->AddOutputLayer(0, "output");
Connect(SpaceToDepth, output, outputTensorInfo, 0, 0);
return net;
}
void SpaceToDepthEndToEnd(const std::vector<armnn::BackendId>& backends,
const armnn::DataLayout& dataLayout,
TensorInfo& inputTensorInfo,
TensorInfo& outputTensorInfo,
std::vector<float>& inputData,
std::vector<float>& expectedOutputData,
const unsigned int blockSize)
{
using namespace armnn;
if (dataLayout == armnn::DataLayout::NCHW){
PermuteDataToNCHW<armnn::DataType::Float32>(backends, dataLayout, inputTensorInfo, inputData);
PermuteDataToNCHW<armnn::DataType::Float32>(backends, dataLayout, outputTensorInfo, expectedOutputData);
}
// Builds up the structure of the network
INetworkPtr net = CreateSpaceToDepthNetwork<armnn::DataType::Float32>(inputTensorInfo.GetShape(),
outputTensorInfo.GetShape(),
dataLayout,
blockSize);
BOOST_TEST_CHECKPOINT("Create a network");
std::map<int, std::vector<float>> inputTensorData = { { 0, inputData } };
std::map<int, std::vector<float>> expectedOutputTensorData = { { 0, expectedOutputData } };
EndToEndLayerTestImpl<armnn::DataType::Float32, armnn::DataType::Float32>(move(net),
inputTensorData,
expectedOutputTensorData,
backends);
}
} // anonymous namespace
|
