| // |
| // 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); |
| } |
| |
| void SpaceToDepthNHWCEndToEndTest1(const std::vector<armnn::BackendId>& defaultBackends) |
| { |
| const unsigned int blockSize = 2; |
| |
| armnn::TensorShape inputShape{1, 2, 2, 1}; |
| armnn::TensorInfo inputTensorInfo(inputShape, armnn::DataType::Float32); |
| |
| armnn::TensorShape outputShape{1, 1, 1, 4}; |
| armnn::TensorInfo outputTensorInfo(outputShape, armnn::DataType::Float32); |
| |
| std::vector<float> inputData = std::vector<float>( |
| { |
| 1.0f, 2.0f, 3.0f, 4.0f |
| }); |
| |
| std::vector<float> expectedOutputData = std::vector<float>( |
| { |
| 1.0f, 2.0f, 3.0f, 4.0f |
| }); |
| |
| SpaceToDepthEndToEnd(defaultBackends, |
| armnn::DataLayout::NHWC, |
| inputTensorInfo, |
| outputTensorInfo, |
| inputData, |
| expectedOutputData, |
| blockSize); |
| } |
| |
| void SpaceToDepthNCHWEndToEndTest1(const std::vector<armnn::BackendId>& defaultBackends) |
| { |
| const unsigned int blockSize = 2; |
| |
| armnn::TensorShape inputShape{1, 2, 2, 1}; |
| armnn::TensorInfo inputTensorInfo(inputShape, armnn::DataType::Float32); |
| |
| armnn::TensorShape outputShape{1, 1, 1, 4}; |
| armnn::TensorInfo outputTensorInfo(outputShape, armnn::DataType::Float32); |
| |
| std::vector<float> inputData = std::vector<float>( |
| { |
| 1.0f, 2.0f, 3.0f, 4.0f |
| }); |
| |
| std::vector<float> expectedOutputData = std::vector<float>( |
| { |
| 1.0f, 2.0f, 3.0f, 4.0f |
| }); |
| |
| SpaceToDepthEndToEnd(defaultBackends, |
| armnn::DataLayout::NCHW, |
| inputTensorInfo, |
| outputTensorInfo, |
| inputData, |
| expectedOutputData, |
| blockSize); |
| } |
| |
| void SpaceToDepthNHWCEndToEndTest2(const std::vector<armnn::BackendId>& defaultBackends) |
| { |
| const unsigned int blockSize = 2; |
| |
| armnn::TensorShape inputShape{1, 2, 2, 2}; |
| armnn::TensorShape outputShape{1, 1, 1, 8}; |
| |
| armnn::TensorInfo outputTensorInfo(outputShape, armnn::DataType::Float32); |
| armnn::TensorInfo inputTensorInfo(inputShape, armnn::DataType::Float32); |
| |
| std::vector<float> inputData = std::vector<float>( |
| { |
| 1.4f, 2.3f, 3.2f, 4.1f, 5.4f, 6.3f, 7.2f, 8.1f |
| }); |
| |
| std::vector<float> expectedOutputData = std::vector<float>( |
| { |
| 1.4f, 2.3f, 3.2f, 4.1f, 5.4f, 6.3f, 7.2f, 8.1f |
| }); |
| |
| SpaceToDepthEndToEnd(defaultBackends, |
| armnn::DataLayout::NHWC, |
| inputTensorInfo, |
| outputTensorInfo, |
| inputData, |
| expectedOutputData, |
| blockSize); |
| } |
| |
| void SpaceToDepthNCHWEndToEndTest2(const std::vector<armnn::BackendId>& defaultBackends) |
| { |
| const unsigned int blockSize = 2; |
| |
| armnn::TensorShape inputShape{1, 2, 2, 2}; |
| armnn::TensorShape outputShape{1, 1, 1, 8}; |
| |
| armnn::TensorInfo inputTensorInfo(inputShape, armnn::DataType::Float32); |
| armnn::TensorInfo outputTensorInfo(outputShape, armnn::DataType::Float32); |
| |
| |
| std::vector<float> inputData = std::vector<float>( |
| { |
| 1.4f, 2.3f, 3.2f, 4.1f, 5.4f, 6.3f, 7.2f, 8.1f |
| }); |
| |
| std::vector<float> expectedOutputData = std::vector<float>( |
| { |
| 1.4f, 2.3f, 3.2f, 4.1f, 5.4f, 6.3f, 7.2f, 8.1f |
| }); |
| |
| SpaceToDepthEndToEnd(defaultBackends, |
| armnn::DataLayout::NCHW, |
| inputTensorInfo, |
| outputTensorInfo, |
| inputData, |
| expectedOutputData, |
| blockSize); |
| } |
| |
| } // anonymous namespace |