diff options
Diffstat (limited to 'src/armnn/test')
-rw-r--r-- | src/armnn/test/CreateWorkload.hpp | 116 | ||||
-rw-r--r-- | src/armnn/test/GraphTests.cpp | 4 | ||||
-rw-r--r-- | src/armnn/test/OptimizerTests.cpp | 56 | ||||
-rw-r--r-- | src/armnn/test/ShapeInferenceTests.cpp | 84 | ||||
-rw-r--r-- | src/armnn/test/SubgraphViewTests.cpp | 2 | ||||
-rw-r--r-- | src/armnn/test/optimizations/AddBroadcastReshapeLayerTests.cpp | 4 | ||||
-rw-r--r-- | src/armnn/test/optimizations/ConvertConstantsBFloatTests.cpp | 4 | ||||
-rw-r--r-- | src/armnn/test/optimizations/ConvertConstantsFloatToHalfTests.cpp | 2 | ||||
-rw-r--r-- | src/armnn/test/optimizations/ConvertConstantsHalfToFloatTests.cpp | 2 | ||||
-rw-r--r-- | src/armnn/test/optimizations/Fp32NetworkToBf16ConverterTests.cpp | 8 |
10 files changed, 141 insertions, 141 deletions
diff --git a/src/armnn/test/CreateWorkload.hpp b/src/armnn/test/CreateWorkload.hpp index 3ea2c35061..12623e62a0 100644 --- a/src/armnn/test/CreateWorkload.hpp +++ b/src/armnn/test/CreateWorkload.hpp @@ -14,9 +14,9 @@ #include <armnn/utility/IgnoreUnused.hpp> #include <armnn/utility/PolymorphicDowncast.hpp> +#include <backendsCommon/TensorHandle.hpp> #include <backendsCommon/WorkloadData.hpp> #include <backendsCommon/WorkloadFactory.hpp> -#include <backendsCommon/CpuTensorHandle.hpp> #include <boost/test/unit_test.hpp> @@ -353,10 +353,10 @@ std::unique_ptr<BatchNormalizationWorkloadType> CreateBatchNormalizationWorkload BatchNormalizationLayer* const layer = graph.AddLayer<BatchNormalizationLayer>(layerDesc, "layer"); armnn::TensorInfo weightInfo({3}, DataType); - layer->m_Mean = std::make_unique<ScopedCpuTensorHandle>(weightInfo); - layer->m_Variance = std::make_unique<ScopedCpuTensorHandle>(weightInfo); - layer->m_Beta = std::make_unique<ScopedCpuTensorHandle>(weightInfo); - layer->m_Gamma = std::make_unique<ScopedCpuTensorHandle>(weightInfo); + layer->m_Mean = std::make_unique<ScopedTensorHandle>(weightInfo); + layer->m_Variance = std::make_unique<ScopedTensorHandle>(weightInfo); + layer->m_Beta = std::make_unique<ScopedTensorHandle>(weightInfo); + layer->m_Gamma = std::make_unique<ScopedTensorHandle>(weightInfo); layer->m_Mean->Allocate(); layer->m_Variance->Allocate(); layer->m_Beta->Allocate(); @@ -411,10 +411,10 @@ std::unique_ptr<BatchNormalizationWorkloadType> CreateBatchNormalizationWithBlob BatchNormalizationLayer* const layer = graph.AddLayer<BatchNormalizationLayer>(layerDesc, "layer"); armnn::TensorInfo weightInfo({3}, DataType); - layer->m_Mean = std::make_unique<ScopedCpuTensorHandle>(weightInfo); - layer->m_Variance = std::make_unique<ScopedCpuTensorHandle>(weightInfo); - layer->m_Beta = std::make_unique<ScopedCpuTensorHandle>(weightInfo); - layer->m_Gamma = std::make_unique<ScopedCpuTensorHandle>(weightInfo); + layer->m_Mean = std::make_unique<ScopedTensorHandle>(weightInfo); + layer->m_Variance = std::make_unique<ScopedTensorHandle>(weightInfo); + layer->m_Beta = std::make_unique<ScopedTensorHandle>(weightInfo); + layer->m_Gamma = std::make_unique<ScopedTensorHandle>(weightInfo); layer->m_Mean->Allocate(); layer->m_Variance->Allocate(); layer->m_Beta->Allocate(); @@ -492,8 +492,8 @@ std::unique_ptr<Convolution2dWorkload> CreateConvolution2dWorkloadTest(armnn::IW TensorShape inputShape = (dataLayout == DataLayout::NCHW) ? TensorShape{2, 3, 8, 16} : TensorShape{2, 8, 16, 3}; TensorShape outputShape = (dataLayout == DataLayout::NCHW) ? TensorShape{2, 2, 2, 10} : TensorShape{2, 2, 10, 2}; - layer->m_Weight = std::make_unique<ScopedCpuTensorHandle>(TensorInfo(weightShape, DataType)); - layer->m_Bias = std::make_unique<ScopedCpuTensorHandle>(TensorInfo({2}, GetBiasDataType(DataType))); + layer->m_Weight = std::make_unique<ScopedTensorHandle>(TensorInfo(weightShape, DataType)); + layer->m_Bias = std::make_unique<ScopedTensorHandle>(TensorInfo({2}, GetBiasDataType(DataType))); layer->m_Weight->Allocate(); layer->m_Bias->Allocate(); @@ -555,8 +555,8 @@ std::unique_ptr<Convolution2dWorkload> CreateConvolution2dFusedActivationWithBlo TensorShape inputShape = (dataLayout == DataLayout::NCHW) ? TensorShape{2, 3, 8, 16} : TensorShape{2, 8, 16, 3}; TensorShape outputShape = (dataLayout == DataLayout::NCHW) ? TensorShape{2, 2, 2, 10} : TensorShape{2, 2, 10, 2}; - layer->m_Weight = std::make_unique<ScopedCpuTensorHandle>(TensorInfo(weightShape, DataType)); - layer->m_Bias = std::make_unique<ScopedCpuTensorHandle>(TensorInfo({2}, GetBiasDataType(DataType))); + layer->m_Weight = std::make_unique<ScopedTensorHandle>(TensorInfo(weightShape, DataType)); + layer->m_Bias = std::make_unique<ScopedTensorHandle>(TensorInfo({2}, GetBiasDataType(DataType))); layer->m_Weight->Allocate(); layer->m_Bias->Allocate(); @@ -639,8 +639,8 @@ std::unique_ptr<Convolution2dWorkload> CreateConvolution2dWorkloadFastMathTest(a TensorShape inputShape = TensorShape{1, 32, 149, 149}; TensorShape outputShape = TensorShape{1, 32, 147, 147}; - layer->m_Weight = std::make_unique<ScopedCpuTensorHandle>(TensorInfo(weightShape, DataType)); - layer->m_Bias = std::make_unique<ScopedCpuTensorHandle>(TensorInfo({2}, GetBiasDataType(DataType))); + layer->m_Weight = std::make_unique<ScopedTensorHandle>(TensorInfo(weightShape, DataType)); + layer->m_Bias = std::make_unique<ScopedTensorHandle>(TensorInfo({2}, GetBiasDataType(DataType))); layer->m_Weight->Allocate(); layer->m_Bias->Allocate(); @@ -692,23 +692,23 @@ std::unique_ptr<LstmWorkload> CreateLstmWorkloadTest(armnn::IWorkloadFactory& fa unsigned int numUnits = 4; unsigned int outputSize = 4; - layer->m_BasicParameters.m_InputToForgetWeights = std::make_unique<ScopedCpuTensorHandle> + layer->m_BasicParameters.m_InputToForgetWeights = std::make_unique<ScopedTensorHandle> (TensorInfo({ numUnits, inputSize }, DataType::Float32)); - layer->m_BasicParameters.m_InputToCellWeights = std::make_unique<ScopedCpuTensorHandle> + layer->m_BasicParameters.m_InputToCellWeights = std::make_unique<ScopedTensorHandle> (TensorInfo({ numUnits, inputSize }, DataType::Float32)); - layer->m_BasicParameters.m_InputToOutputWeights = std::make_unique<ScopedCpuTensorHandle> + layer->m_BasicParameters.m_InputToOutputWeights = std::make_unique<ScopedTensorHandle> (TensorInfo({ numUnits, inputSize }, DataType::Float32)); - layer->m_BasicParameters.m_RecurrentToForgetWeights = std::make_unique<ScopedCpuTensorHandle> + layer->m_BasicParameters.m_RecurrentToForgetWeights = std::make_unique<ScopedTensorHandle> (TensorInfo({ numUnits, outputSize }, DataType::Float32)); - layer->m_BasicParameters.m_RecurrentToCellWeights = std::make_unique<ScopedCpuTensorHandle> + layer->m_BasicParameters.m_RecurrentToCellWeights = std::make_unique<ScopedTensorHandle> (TensorInfo({ numUnits, outputSize }, DataType::Float32)); - layer->m_BasicParameters.m_RecurrentToOutputWeights = std::make_unique<ScopedCpuTensorHandle> + layer->m_BasicParameters.m_RecurrentToOutputWeights = std::make_unique<ScopedTensorHandle> (TensorInfo({ numUnits, outputSize }, DataType::Float32)); - layer->m_BasicParameters.m_ForgetGateBias = std::make_unique<ScopedCpuTensorHandle> + layer->m_BasicParameters.m_ForgetGateBias = std::make_unique<ScopedTensorHandle> (TensorInfo({ numUnits }, DataType::Float32)); - layer->m_BasicParameters.m_CellBias = std::make_unique<ScopedCpuTensorHandle> + layer->m_BasicParameters.m_CellBias = std::make_unique<ScopedTensorHandle> (TensorInfo({ numUnits }, DataType::Float32)); - layer->m_BasicParameters.m_OutputGateBias = std::make_unique<ScopedCpuTensorHandle> + layer->m_BasicParameters.m_OutputGateBias = std::make_unique<ScopedTensorHandle> (TensorInfo({ numUnits }, DataType::Float32)); layer->m_BasicParameters.m_InputToForgetWeights->Allocate(); @@ -724,9 +724,9 @@ std::unique_ptr<LstmWorkload> CreateLstmWorkloadTest(armnn::IWorkloadFactory& fa if (layerDesc.m_PeepholeEnabled) { - layer->m_PeepholeParameters.m_CellToForgetWeights = std::make_unique<ScopedCpuTensorHandle> + layer->m_PeepholeParameters.m_CellToForgetWeights = std::make_unique<ScopedTensorHandle> (TensorInfo({ numUnits }, DataType::Float32)); - layer->m_PeepholeParameters.m_CellToOutputWeights = std::make_unique<ScopedCpuTensorHandle> + layer->m_PeepholeParameters.m_CellToOutputWeights = std::make_unique<ScopedTensorHandle> (TensorInfo({ numUnits }, DataType::Float32)); layer->m_PeepholeParameters.m_CellToForgetWeights->Allocate(); layer->m_PeepholeParameters.m_CellToOutputWeights->Allocate(); @@ -814,27 +814,27 @@ std::unique_ptr<QuantizedLstmWorkload> CreateQuantizedLstmWorkloadTest(armnn::IW // Weights and bias layer->m_QuantizedLstmParameters.m_InputToInputWeights = - std::make_unique<ScopedCpuTensorHandle>(inputWeightsInfo); + std::make_unique<ScopedTensorHandle>(inputWeightsInfo); layer->m_QuantizedLstmParameters.m_InputToForgetWeights = - std::make_unique<ScopedCpuTensorHandle>(inputWeightsInfo); + std::make_unique<ScopedTensorHandle>(inputWeightsInfo); layer->m_QuantizedLstmParameters.m_InputToCellWeights = - std::make_unique<ScopedCpuTensorHandle>(inputWeightsInfo); + std::make_unique<ScopedTensorHandle>(inputWeightsInfo); layer->m_QuantizedLstmParameters.m_InputToOutputWeights = - std::make_unique<ScopedCpuTensorHandle>(inputWeightsInfo); + std::make_unique<ScopedTensorHandle>(inputWeightsInfo); layer->m_QuantizedLstmParameters.m_RecurrentToInputWeights = - std::make_unique<ScopedCpuTensorHandle>(recurrentWeightsInfo); + std::make_unique<ScopedTensorHandle>(recurrentWeightsInfo); layer->m_QuantizedLstmParameters.m_RecurrentToForgetWeights = - std::make_unique<ScopedCpuTensorHandle>(recurrentWeightsInfo); + std::make_unique<ScopedTensorHandle>(recurrentWeightsInfo); layer->m_QuantizedLstmParameters.m_RecurrentToCellWeights = - std::make_unique<ScopedCpuTensorHandle>(recurrentWeightsInfo); + std::make_unique<ScopedTensorHandle>(recurrentWeightsInfo); layer->m_QuantizedLstmParameters.m_RecurrentToOutputWeights = - std::make_unique<ScopedCpuTensorHandle>(recurrentWeightsInfo); + std::make_unique<ScopedTensorHandle>(recurrentWeightsInfo); - layer->m_QuantizedLstmParameters.m_InputGateBias = std::make_unique<ScopedCpuTensorHandle>(biasInfo); - layer->m_QuantizedLstmParameters.m_ForgetGateBias = std::make_unique<ScopedCpuTensorHandle>(biasInfo); - layer->m_QuantizedLstmParameters.m_CellBias = std::make_unique<ScopedCpuTensorHandle>(biasInfo); - layer->m_QuantizedLstmParameters.m_OutputGateBias = std::make_unique<ScopedCpuTensorHandle>(biasInfo); + layer->m_QuantizedLstmParameters.m_InputGateBias = std::make_unique<ScopedTensorHandle>(biasInfo); + layer->m_QuantizedLstmParameters.m_ForgetGateBias = std::make_unique<ScopedTensorHandle>(biasInfo); + layer->m_QuantizedLstmParameters.m_CellBias = std::make_unique<ScopedTensorHandle>(biasInfo); + layer->m_QuantizedLstmParameters.m_OutputGateBias = std::make_unique<ScopedTensorHandle>(biasInfo); // Allocate weights and bias layer->m_QuantizedLstmParameters.m_InputToInputWeights->Allocate(); @@ -977,27 +977,27 @@ std::unique_ptr<QLstmWorkload> CreateQLstmWorkloadTest(armnn::IWorkloadFactory& armnn::TensorInfo layerNormWeightsInfo({numUnits}, armnn::DataType::QSymmS16, layerNormScale, layerNormOffset); // Create and allocate tensors - layer->m_BasicParameters.m_InputToForgetWeights = std::make_unique<ScopedCpuTensorHandle>(inputWeightsInfo); - layer->m_BasicParameters.m_InputToCellWeights = std::make_unique<ScopedCpuTensorHandle>(inputWeightsInfo); - layer->m_BasicParameters.m_InputToOutputWeights = std::make_unique<ScopedCpuTensorHandle>(inputWeightsInfo); + layer->m_BasicParameters.m_InputToForgetWeights = std::make_unique<ScopedTensorHandle>(inputWeightsInfo); + layer->m_BasicParameters.m_InputToCellWeights = std::make_unique<ScopedTensorHandle>(inputWeightsInfo); + layer->m_BasicParameters.m_InputToOutputWeights = std::make_unique<ScopedTensorHandle>(inputWeightsInfo); layer->m_BasicParameters.m_RecurrentToForgetWeights = - std::make_unique<ScopedCpuTensorHandle>(recurrentWeightsInfo); + std::make_unique<ScopedTensorHandle>(recurrentWeightsInfo); layer->m_BasicParameters.m_RecurrentToCellWeights = - std::make_unique<ScopedCpuTensorHandle>(recurrentWeightsInfo); + std::make_unique<ScopedTensorHandle>(recurrentWeightsInfo); layer->m_BasicParameters.m_RecurrentToOutputWeights = - std::make_unique<ScopedCpuTensorHandle>(recurrentWeightsInfo); + std::make_unique<ScopedTensorHandle>(recurrentWeightsInfo); - layer->m_BasicParameters.m_ForgetGateBias = std::make_unique<ScopedCpuTensorHandle>(biasInfo); - layer->m_BasicParameters.m_CellBias = std::make_unique<ScopedCpuTensorHandle>(biasInfo); - layer->m_BasicParameters.m_OutputGateBias = std::make_unique<ScopedCpuTensorHandle>(biasInfo); + layer->m_BasicParameters.m_ForgetGateBias = std::make_unique<ScopedTensorHandle>(biasInfo); + layer->m_BasicParameters.m_CellBias = std::make_unique<ScopedTensorHandle>(biasInfo); + layer->m_BasicParameters.m_OutputGateBias = std::make_unique<ScopedTensorHandle>(biasInfo); layer->m_LayerNormParameters.m_ForgetLayerNormWeights = - std::make_unique<ScopedCpuTensorHandle>(layerNormWeightsInfo); + std::make_unique<ScopedTensorHandle>(layerNormWeightsInfo); layer->m_LayerNormParameters.m_CellLayerNormWeights = - std::make_unique<ScopedCpuTensorHandle>(layerNormWeightsInfo); + std::make_unique<ScopedTensorHandle>(layerNormWeightsInfo); layer->m_LayerNormParameters.m_OutputLayerNormWeights = - std::make_unique<ScopedCpuTensorHandle>(layerNormWeightsInfo); + std::make_unique<ScopedTensorHandle>(layerNormWeightsInfo); layer->m_BasicParameters.m_InputToForgetWeights->Allocate(); layer->m_BasicParameters.m_InputToCellWeights->Allocate(); @@ -1093,8 +1093,8 @@ std::unique_ptr<Convolution2dWorkload> CreateDirectConvolution2dWorkloadTest(arm float inputsQScale = DataType == armnn::DataType::QAsymmU8 ? 1.0f : 0.0; float outputQScale = DataType == armnn::DataType::QAsymmU8 ? 2.0f : 0.0; - layer->m_Weight = std::make_unique<ScopedCpuTensorHandle>(TensorInfo({ 2, 3, 3, 3 }, DataType, inputsQScale)); - layer->m_Bias = std::make_unique<ScopedCpuTensorHandle> + layer->m_Weight = std::make_unique<ScopedTensorHandle>(TensorInfo({ 2, 3, 3, 3 }, DataType, inputsQScale)); + layer->m_Bias = std::make_unique<ScopedTensorHandle> (TensorInfo({2}, GetBiasDataType(DataType), inputsQScale)); layer->m_Weight->Allocate(); layer->m_Bias->Allocate(); @@ -1148,7 +1148,7 @@ std::unique_ptr<DepthwiseConvolution2dFloat32Workload> CreateDepthwiseConvolutio DepthwiseConvolution2dLayer* const layer = graph.AddLayer<DepthwiseConvolution2dLayer>(layerDesc, "layer"); - layer->m_Weight = std::make_unique<ScopedCpuTensorHandle>(TensorInfo({1, 2, 4, 4}, DataType)); // [ M, I, H, W ] + layer->m_Weight = std::make_unique<ScopedTensorHandle>(TensorInfo({1, 2, 4, 4}, DataType)); // [ M, I, H, W ] layer->m_Weight->Allocate(); // Creates extra layers. @@ -1200,8 +1200,8 @@ std::unique_ptr<FullyConnectedWorkload> CreateFullyConnectedWorkloadTest(armnn:: float inputsQScale = DataType == armnn::DataType::QAsymmU8 ? 1.0f : 0.0; float outputQScale = DataType == armnn::DataType::QAsymmU8 ? 2.0f : 0.0; - layer->m_Weight = std::make_unique<ScopedCpuTensorHandle>(TensorInfo({7, 20}, DataType, inputsQScale, 0)); - layer->m_Bias = std::make_unique<ScopedCpuTensorHandle>(TensorInfo({7}, GetBiasDataType(DataType), inputsQScale)); + layer->m_Weight = std::make_unique<ScopedTensorHandle>(TensorInfo({7, 20}, DataType, inputsQScale, 0)); + layer->m_Bias = std::make_unique<ScopedTensorHandle>(TensorInfo({7}, GetBiasDataType(DataType), inputsQScale)); layer->m_Weight->Allocate(); layer->m_Bias->Allocate(); @@ -1245,8 +1245,8 @@ std::unique_ptr<FullyConnectedWorkload> CreateFullyConnectedWithBlobWorkloadTest float inputsQScale = DataType == armnn::DataType::QAsymmU8 ? 1.0f : 0.0; float outputQScale = DataType == armnn::DataType::QAsymmU8 ? 2.0f : 0.0; - layer->m_Weight = std::make_unique<ScopedCpuTensorHandle>(TensorInfo({7, 20}, DataType, inputsQScale, 0)); - layer->m_Bias = std::make_unique<ScopedCpuTensorHandle>(TensorInfo({7}, GetBiasDataType(DataType), inputsQScale)); + layer->m_Weight = std::make_unique<ScopedTensorHandle>(TensorInfo({7, 20}, DataType, inputsQScale, 0)); + layer->m_Bias = std::make_unique<ScopedTensorHandle>(TensorInfo({7}, GetBiasDataType(DataType), inputsQScale)); layer->m_Weight->Allocate(); layer->m_Bias->Allocate(); @@ -2108,7 +2108,7 @@ std::unique_ptr<ConstantWorkload> CreateConstantWorkloadTest(armnn::IWorkloadFac armnn::TensorInfo outputTensorInfo(outputShape, DataType); auto constant = graph.AddLayer<ConstantLayer>("constant"); - constant->m_LayerOutput = std::make_unique<ScopedCpuTensorHandle>(outputTensorInfo); + constant->m_LayerOutput = std::make_unique<ScopedTensorHandle>(outputTensorInfo); BOOST_TEST_CHECKPOINT("created constant layer"); Layer* const output = graph.AddLayer<OutputLayer>(0, "output"); diff --git a/src/armnn/test/GraphTests.cpp b/src/armnn/test/GraphTests.cpp index 602575b352..69f96d43a3 100644 --- a/src/armnn/test/GraphTests.cpp +++ b/src/armnn/test/GraphTests.cpp @@ -14,7 +14,7 @@ #include <armnn/backends/IBackendInternal.hpp> -#include <backendsCommon/CpuTensorHandle.hpp> +#include <backendsCommon/TensorHandle.hpp> #include <backendsCommon/TensorHandleFactoryRegistry.hpp> #include <boost/test/unit_test.hpp> @@ -603,7 +603,7 @@ BOOST_AUTO_TEST_CASE(CheckGraphConstTensorSharing) float weight = 1.0f; armnn::ConstTensor constTensor({{ 1, 1 }, armnn::DataType::Float32}, &weight); - fcLayer->m_Weight = std::make_shared<armnn::ScopedCpuTensorHandle>(constTensor);; + fcLayer->m_Weight = std::make_shared<armnn::ScopedTensorHandle>(constTensor);; // point sharedWeightPtr to graph1's const tensor sharedWeightPtr = fcLayer->m_Weight->GetConstTensor<float>(); diff --git a/src/armnn/test/OptimizerTests.cpp b/src/armnn/test/OptimizerTests.cpp index 7e8a898293..d0734d83be 100644 --- a/src/armnn/test/OptimizerTests.cpp +++ b/src/armnn/test/OptimizerTests.cpp @@ -18,9 +18,9 @@ #include <armnn/utility/PolymorphicDowncast.hpp> #include <armnnUtils/FloatingPointConverter.hpp> -#include <backendsCommon/CpuTensorHandle.hpp> #include <backendsCommon/IBackendInternal.hpp> #include <backendsCommon/LayerSupportBase.hpp> +#include <backendsCommon/TensorHandle.hpp> #include <boost/test/unit_test.hpp> @@ -45,23 +45,23 @@ void CreateLSTMLayerHelper(Graph &graph, bool CifgEnabled) unsigned int numUnits = 4; unsigned int outputSize = 4; - layer->m_BasicParameters.m_InputToForgetWeights = std::make_unique<ScopedCpuTensorHandle> + layer->m_BasicParameters.m_InputToForgetWeights = std::make_unique<ScopedTensorHandle> (TensorInfo({ numUnits, inputSize }, DataType::Float32)); - layer->m_BasicParameters.m_InputToCellWeights = std::make_unique<ScopedCpuTensorHandle> + layer->m_BasicParameters.m_InputToCellWeights = std::make_unique<ScopedTensorHandle> (TensorInfo({ numUnits, inputSize }, DataType::Float32)); - layer->m_BasicParameters.m_InputToOutputWeights = std::make_unique<ScopedCpuTensorHandle> + layer->m_BasicParameters.m_InputToOutputWeights = std::make_unique<ScopedTensorHandle> (TensorInfo({ numUnits, inputSize }, DataType::Float32)); - layer->m_BasicParameters.m_RecurrentToForgetWeights = std::make_unique<ScopedCpuTensorHandle> + layer->m_BasicParameters.m_RecurrentToForgetWeights = std::make_unique<ScopedTensorHandle> (TensorInfo({ numUnits, outputSize }, DataType::Float32)); - layer->m_BasicParameters.m_RecurrentToCellWeights = std::make_unique<ScopedCpuTensorHandle> + layer->m_BasicParameters.m_RecurrentToCellWeights = std::make_unique<ScopedTensorHandle> (TensorInfo({ numUnits, outputSize }, DataType::Float32)); - layer->m_BasicParameters.m_RecurrentToOutputWeights = std::make_unique<ScopedCpuTensorHandle> + layer->m_BasicParameters.m_RecurrentToOutputWeights = std::make_unique<ScopedTensorHandle> (TensorInfo({ numUnits, outputSize }, DataType::Float32)); - layer->m_BasicParameters.m_ForgetGateBias = std::make_unique<ScopedCpuTensorHandle> + layer->m_BasicParameters.m_ForgetGateBias = std::make_unique<ScopedTensorHandle> (TensorInfo({ numUnits }, DataType::Float32)); - layer->m_BasicParameters.m_CellBias = std::make_unique<ScopedCpuTensorHandle> + layer->m_BasicParameters.m_CellBias = std::make_unique<ScopedTensorHandle> (TensorInfo({ numUnits }, DataType::Float32)); - layer->m_BasicParameters.m_OutputGateBias = std::make_unique<ScopedCpuTensorHandle> + layer->m_BasicParameters.m_OutputGateBias = std::make_unique<ScopedTensorHandle> (TensorInfo({ numUnits }, DataType::Float32)); layer->m_BasicParameters.m_InputToForgetWeights->Allocate(); @@ -76,11 +76,11 @@ void CreateLSTMLayerHelper(Graph &graph, bool CifgEnabled) if (!layerDesc.m_CifgEnabled) { - layer->m_CifgParameters.m_InputToInputWeights = std::make_unique<ScopedCpuTensorHandle> + layer->m_CifgParameters.m_InputToInputWeights = std::make_unique<ScopedTensorHandle> (TensorInfo({ numUnits, inputSize }, DataType::Float32)); - layer->m_CifgParameters.m_RecurrentToInputWeights = std::make_unique<ScopedCpuTensorHandle> + layer->m_CifgParameters.m_RecurrentToInputWeights = std::make_unique<ScopedTensorHandle> (TensorInfo({ numUnits, outputSize }, DataType::Float32)); - layer->m_CifgParameters.m_InputGateBias = std::make_unique<ScopedCpuTensorHandle> + layer->m_CifgParameters.m_InputGateBias = std::make_unique<ScopedTensorHandle> (TensorInfo({ numUnits }, DataType::Float32)); layer->m_CifgParameters.m_InputToInputWeights->Allocate(); layer->m_CifgParameters.m_RecurrentToInputWeights->Allocate(); @@ -89,9 +89,9 @@ void CreateLSTMLayerHelper(Graph &graph, bool CifgEnabled) if (layerDesc.m_ProjectionEnabled) { - layer->m_ProjectionParameters.m_ProjectionWeights = std::make_unique<ScopedCpuTensorHandle> + layer->m_ProjectionParameters.m_ProjectionWeights = std::make_unique<ScopedTensorHandle> (TensorInfo({ outputSize, numUnits }, DataType::Float32)); - layer->m_ProjectionParameters.m_ProjectionBias = std::make_unique<ScopedCpuTensorHandle> + layer->m_ProjectionParameters.m_ProjectionBias = std::make_unique<ScopedTensorHandle> (TensorInfo({ outputSize }, DataType::Float32)); layer->m_ProjectionParameters.m_ProjectionWeights->Allocate(); layer->m_ProjectionParameters.m_ProjectionBias->Allocate(); @@ -101,13 +101,13 @@ void CreateLSTMLayerHelper(Graph &graph, bool CifgEnabled) { if (!layerDesc.m_CifgEnabled) { - layer->m_PeepholeParameters.m_CellToInputWeights = std::make_unique<ScopedCpuTensorHandle> + layer->m_PeepholeParameters.m_CellToInputWeights = std::make_unique<ScopedTensorHandle> (TensorInfo({ numUnits }, DataType::Float32)); layer->m_PeepholeParameters.m_CellToInputWeights->Allocate(); } - layer->m_PeepholeParameters.m_CellToForgetWeights = std::make_unique<ScopedCpuTensorHandle> + layer->m_PeepholeParameters.m_CellToForgetWeights = std::make_unique<ScopedTensorHandle> (TensorInfo({ numUnits }, DataType::Float32)); - layer->m_PeepholeParameters.m_CellToOutputWeights = std::make_unique<ScopedCpuTensorHandle> + layer->m_PeepholeParameters.m_CellToOutputWeights = std::make_unique<ScopedTensorHandle> (TensorInfo({ numUnits }, DataType::Float32)); layer->m_PeepholeParameters.m_CellToForgetWeights->Allocate(); layer->m_PeepholeParameters.m_CellToOutputWeights->Allocate(); @@ -276,7 +276,7 @@ void CreateConvolution2dGraph(Graph &graph, const unsigned int* inputShape, input->GetOutputSlot().SetTensorInfo(inputInfo); Convolution2dLayer* layer = graph.AddLayer<Convolution2dLayer>(desc, "conv2d"); - layer->m_Weight = std::make_unique<armnn::ScopedCpuTensorHandle>(weights); + layer->m_Weight = std::make_unique<armnn::ScopedTensorHandle>(weights); layer->GetOutputSlot().SetTensorInfo(outputInfo); Layer* output = graph.AddLayer<OutputLayer>(0, "output"); @@ -326,7 +326,7 @@ void CreateDepthwiseConvolution2dGraph(Graph &graph, const unsigned int* inputSh input->GetOutputSlot().SetTensorInfo(inputInfo); DepthwiseConvolution2dLayer* layer = graph.AddLayer<DepthwiseConvolution2dLayer>(desc, "depthwiseConv2d"); - layer->m_Weight = std::make_unique<armnn::ScopedCpuTensorHandle>(weights); + layer->m_Weight = std::make_unique<armnn::ScopedTensorHandle>(weights); layer->GetOutputSlot().SetTensorInfo(outputInfo); Layer* output = graph.AddLayer<OutputLayer>(0, "output"); @@ -529,7 +529,7 @@ BOOST_AUTO_TEST_CASE(DetectionPostProcessValidateTensorShapes) descriptor.m_MaxDetections = 3; DetectionPostProcessLayer* layer = graph.AddLayer<DetectionPostProcessLayer>(descriptor, "detectionPostProcess"); - layer->m_Anchors = std::make_unique<armnn::ScopedCpuTensorHandle>(anchors); + layer->m_Anchors = std::make_unique<armnn::ScopedTensorHandle>(anchors); layer->GetOutputSlot(0).SetTensorInfo(detectionBoxesInfo); layer->GetOutputSlot(1).SetTensorInfo(detectionScoresInfo); layer->GetOutputSlot(2).SetTensorInfo(detectionClassesInfo); @@ -571,7 +571,7 @@ BOOST_AUTO_TEST_CASE(FoldPadLayerIntoConvolution2dLayer) armnn::ConstTensor weights(armnn::TensorInfo(4, weightsShape, armnn::DataType::Float32), weightsVector); Convolution2dLayer* conv2dLayer = graph.AddLayer<Convolution2dLayer>(convolution2dDescriptor, "conv2d"); - conv2dLayer->m_Weight = std::make_unique<armnn::ScopedCpuTensorHandle>(weights); + conv2dLayer->m_Weight = std::make_unique<armnn::ScopedTensorHandle>(weights); conv2dLayer->GetOutputSlot().SetTensorInfo(outputInfo); Layer* output = graph.AddLayer<OutputLayer>(0, "output"); @@ -1211,16 +1211,16 @@ BOOST_AUTO_TEST_CASE(OptimizeForExclusiveConnectionsFuseTest) input->GetOutputSlot().SetTensorInfo(inputInfo); conv->GetOutputSlot().SetTensorInfo(outputInfo); batchNorm->GetOutputSlot().SetTensorInfo(outputInfo); - conv->m_Weight = std::make_unique<ScopedCpuTensorHandle>(weights); - batchNorm->m_Beta = std::make_unique<ScopedCpuTensorHandle>(beta); - batchNorm->m_Gamma = std::make_unique<ScopedCpuTensorHandle>(gamma); - batchNorm->m_Mean = std::make_unique<ScopedCpuTensorHandle>(mean); - batchNorm->m_Variance = std::make_unique<ScopedCpuTensorHandle>(variance); + conv->m_Weight = std::make_unique<ScopedTensorHandle>(weights); + batchNorm->m_Beta = std::make_unique<ScopedTensorHandle>(beta); + batchNorm->m_Gamma = std::make_unique<ScopedTensorHandle>(gamma); + batchNorm->m_Mean = std::make_unique<ScopedTensorHandle>(mean); + batchNorm->m_Variance = std::make_unique<ScopedTensorHandle>(variance); if (convolution2dDescriptor.m_BiasEnabled) { std::vector<float> biasVector = { 11 }; ConstTensor bias(TensorInfo(1, outputChannelSize, DataType::Float32), biasVector); - conv->m_Bias = std::make_unique<ScopedCpuTensorHandle>(bias); + conv->m_Bias = std::make_unique<ScopedTensorHandle>(bias); } // Connect layers diff --git a/src/armnn/test/ShapeInferenceTests.cpp b/src/armnn/test/ShapeInferenceTests.cpp index 25b0feaded..fa3f400569 100644 --- a/src/armnn/test/ShapeInferenceTests.cpp +++ b/src/armnn/test/ShapeInferenceTests.cpp @@ -9,8 +9,8 @@ #include <Graph.hpp> #include <InternalTypes.hpp> #include <layers/FullyConnectedLayer.hpp> +#include <backendsCommon/TensorHandle.hpp> #include <backendsCommon/WorkloadData.hpp> -#include <backendsCommon/CpuTensorHandle.hpp> #include <string> @@ -240,7 +240,7 @@ BOOST_AUTO_TEST_CASE(ConstantTesst) const float Datum = 0.0f; ConstTensor output0({outputShape, DataType::Float32}, &Datum); - layer->m_LayerOutput = std::make_unique<ScopedCpuTensorHandle>(output0); + layer->m_LayerOutput = std::make_unique<ScopedTensorHandle>(output0); layer->GetOutputSlot(0).SetTensorInfo({{1, 1, 3, 3}, DataType::Float32}); @@ -294,7 +294,7 @@ BOOST_AUTO_TEST_CASE(Convolution2dTest) const float Datum = 0.0f; ConstTensor weights({{1, 1, 3, 3}, DataType::Float32}, &Datum); - layer->m_Weight = std::make_unique<ScopedCpuTensorHandle>(weights); + layer->m_Weight = std::make_unique<ScopedTensorHandle>(weights); RunShapeInferenceTest<Convolution2dLayer>(layer, {{ 1, 1, 4, 4 }}); } @@ -339,7 +339,7 @@ BOOST_AUTO_TEST_CASE(DepthwiseConvolutionTest) const float Datum = 0.0f; ConstTensor weights({{ 2, 5, 3, 2 }, DataType::Float32}, &Datum); - layer->m_Weight = std::make_unique<ScopedCpuTensorHandle>(weights); + layer->m_Weight = std::make_unique<ScopedTensorHandle>(weights); RunShapeInferenceTest<DepthwiseConvolution2dLayer>(layer, {{ 8, 18, 1, 2 }}); } @@ -379,7 +379,7 @@ BOOST_AUTO_TEST_CASE(DetectionPostProcessTest) descriptor, "detectionpostprocess"); - layer->m_Anchors = std::make_unique<ScopedCpuTensorHandle>(anchorsTensor); + layer->m_Anchors = std::make_unique<ScopedTensorHandle>(anchorsTensor); RunShapeInferenceTest<DetectionPostProcessLayer>(layer, {{ 1, 3, 4 }, { 1, 3 }, { 1, 3 }, { 1 }}); } @@ -415,7 +415,7 @@ BOOST_AUTO_TEST_CASE(FullyConnectedTest) const float Datum = 0.0f; ConstTensor weights({{inputChannels, outputChannels}, DataType::Float32}, &Datum); - layer->m_Weight = std::make_unique<ScopedCpuTensorHandle>(weights); + layer->m_Weight = std::make_unique<ScopedTensorHandle>(weights); RunShapeInferenceTest<FullyConnectedLayer>(layer, {{ 1, outputChannels }}); } @@ -469,18 +469,18 @@ BOOST_AUTO_TEST_CASE(LstmTest) float Datum = 0.0f; ConstTensor constTensor({{ 2, 5, 3, 2 }, DataType::Float32}, &Datum); - layer->m_BasicParameters.m_InputToCellWeights = std::make_unique<ScopedCpuTensorHandle>(constTensor); - layer->m_BasicParameters.m_InputToForgetWeights = std::make_unique<ScopedCpuTensorHandle>(constTensor); - layer->m_BasicParameters.m_CellBias = std::make_unique<ScopedCpuTensorHandle>(constTensor); - layer->m_BasicParameters.m_ForgetGateBias = std::make_unique<ScopedCpuTensorHandle>(constTensor); - layer->m_CifgParameters.m_InputGateBias = std::make_unique<ScopedCpuTensorHandle>(constTensor); - layer->m_BasicParameters.m_OutputGateBias = std::make_unique<ScopedCpuTensorHandle>(constTensor); - layer->m_BasicParameters.m_RecurrentToForgetWeights = std::make_unique<ScopedCpuTensorHandle>(constTensor); - layer->m_BasicParameters.m_RecurrentToCellWeights = std::make_unique<ScopedCpuTensorHandle>(constTensor); - layer->m_BasicParameters.m_InputToOutputWeights = std::make_unique<ScopedCpuTensorHandle>(constTensor); - layer->m_BasicParameters.m_RecurrentToOutputWeights = std::make_unique<ScopedCpuTensorHandle>(constTensor); - layer->m_CifgParameters.m_RecurrentToInputWeights = std::make_unique<ScopedCpuTensorHandle>(constTensor); - layer->m_CifgParameters.m_InputToInputWeights = std::make_unique<ScopedCpuTensorHandle>(constTensor); + layer->m_BasicParameters.m_InputToCellWeights = std::make_unique<ScopedTensorHandle>(constTensor); + layer->m_BasicParameters.m_InputToForgetWeights = std::make_unique<ScopedTensorHandle>(constTensor); + layer->m_BasicParameters.m_CellBias = std::make_unique<ScopedTensorHandle>(constTensor); + layer->m_BasicParameters.m_ForgetGateBias = std::make_unique<ScopedTensorHandle>(constTensor); + layer->m_CifgParameters.m_InputGateBias = std::make_unique<ScopedTensorHandle>(constTensor); + layer->m_BasicParameters.m_OutputGateBias = std::make_unique<ScopedTensorHandle>(constTensor); + layer->m_BasicParameters.m_RecurrentToForgetWeights = std::make_unique<ScopedTensorHandle>(constTensor); + layer->m_BasicParameters.m_RecurrentToCellWeights = std::make_unique<ScopedTensorHandle>(constTensor); + layer->m_BasicParameters.m_InputToOutputWeights = std::make_unique<ScopedTensorHandle>(constTensor); + layer->m_BasicParameters.m_RecurrentToOutputWeights = std::make_unique<ScopedTensorHandle>(constTensor); + layer->m_CifgParameters.m_RecurrentToInputWeights = std::make_unique<ScopedTensorHandle>(constTensor); + layer->m_CifgParameters.m_InputToInputWeights = std::make_unique<ScopedTensorHandle>(constTensor); RunShapeInferenceTest<LstmLayer>(layer, {{2, 80}, {2, 20}, {2, 20}, {2, 20}}); } @@ -557,18 +557,18 @@ BOOST_AUTO_TEST_CASE(QLstmTest) float Datum = 0.0f; ConstTensor constTensor({{ 2, 5, 3, 2 }, DataType::Float32}, &Datum); - layer->m_BasicParameters.m_InputToCellWeights = std::make_unique<ScopedCpuTensorHandle>(constTensor); - layer->m_BasicParameters.m_InputToForgetWeights = std::make_unique<ScopedCpuTensorHandle>(constTensor); - layer->m_BasicParameters.m_CellBias = std::make_unique<ScopedCpuTensorHandle>(constTensor); - layer->m_BasicParameters.m_ForgetGateBias = std::make_unique<ScopedCpuTensorHandle>(constTensor); - layer->m_CifgParameters.m_InputGateBias = std::make_unique<ScopedCpuTensorHandle>(constTensor); - layer->m_BasicParameters.m_OutputGateBias = std::make_unique<ScopedCpuTensorHandle>(constTensor); - layer->m_BasicParameters.m_RecurrentToForgetWeights = std::make_unique<ScopedCpuTensorHandle>(constTensor); - layer->m_BasicParameters.m_RecurrentToCellWeights = std::make_unique<ScopedCpuTensorHandle>(constTensor); - layer->m_BasicParameters.m_InputToOutputWeights = std::make_unique<ScopedCpuTensorHandle>(constTensor); - layer->m_BasicParameters.m_RecurrentToOutputWeights = std::make_unique<ScopedCpuTensorHandle>(constTensor); - layer->m_CifgParameters.m_RecurrentToInputWeights = std::make_unique<ScopedCpuTensorHandle>(constTensor); - layer->m_CifgParameters.m_InputToInputWeights = std::make_unique<ScopedCpuTensorHandle>(constTensor); + layer->m_BasicParameters.m_InputToCellWeights = std::make_unique<ScopedTensorHandle>(constTensor); + layer->m_BasicParameters.m_InputToForgetWeights = std::make_unique<ScopedTensorHandle>(constTensor); + layer->m_BasicParameters.m_CellBias = std::make_unique<ScopedTensorHandle>(constTensor); + layer->m_BasicParameters.m_ForgetGateBias = std::make_unique<ScopedTensorHandle>(constTensor); + layer->m_CifgParameters.m_InputGateBias = std::make_unique<ScopedTensorHandle>(constTensor); + layer->m_BasicParameters.m_OutputGateBias = std::make_unique<ScopedTensorHandle>(constTensor); + layer->m_BasicParameters.m_RecurrentToForgetWeights = std::make_unique<ScopedTensorHandle>(constTensor); + layer->m_BasicParameters.m_RecurrentToCellWeights = std::make_unique<ScopedTensorHandle>(constTensor); + layer->m_BasicParameters.m_InputToOutputWeights = std::make_unique<ScopedTensorHandle>(constTensor); + layer->m_BasicParameters.m_RecurrentToOutputWeights = std::make_unique<ScopedTensorHandle>(constTensor); + layer->m_CifgParameters.m_RecurrentToInputWeights = std::make_unique<ScopedTensorHandle>(constTensor); + layer->m_CifgParameters.m_InputToInputWeights = std::make_unique<ScopedTensorHandle>(constTensor); RunShapeInferenceTest<QLstmLayer>(layer, {{2, 20}, {2, 20}, {2, 20}}); } @@ -585,18 +585,18 @@ BOOST_AUTO_TEST_CASE(QuantizedLstmTest) float Datum = 0.0f; ConstTensor constTensor({{ 2, 5, 3, 2 }, DataType::Float32}, &Datum); - layer->m_QuantizedLstmParameters.m_InputToCellWeights = std::make_unique<ScopedCpuTensorHandle>(constTensor); - layer->m_QuantizedLstmParameters.m_InputToForgetWeights = std::make_unique<ScopedCpuTensorHandle>(constTensor); - layer->m_QuantizedLstmParameters.m_CellBias = std::make_unique<ScopedCpuTensorHandle>(constTensor); - layer->m_QuantizedLstmParameters.m_ForgetGateBias = std::make_unique<ScopedCpuTensorHandle>(constTensor); - layer->m_QuantizedLstmParameters.m_InputGateBias = std::make_unique<ScopedCpuTensorHandle>(constTensor); - layer->m_QuantizedLstmParameters.m_OutputGateBias = std::make_unique<ScopedCpuTensorHandle>(constTensor); - layer->m_QuantizedLstmParameters.m_RecurrentToForgetWeights = std::make_unique<ScopedCpuTensorHandle>(constTensor); - layer->m_QuantizedLstmParameters.m_RecurrentToCellWeights = std::make_unique<ScopedCpuTensorHandle>(constTensor); - layer->m_QuantizedLstmParameters.m_InputToOutputWeights = std::make_unique<ScopedCpuTensorHandle>(constTensor); - layer->m_QuantizedLstmParameters.m_RecurrentToOutputWeights = std::make_unique<ScopedCpuTensorHandle>(constTensor); - layer->m_QuantizedLstmParameters.m_RecurrentToInputWeights = std::make_unique<ScopedCpuTensorHandle>(constTensor); - layer->m_QuantizedLstmParameters.m_InputToInputWeights = std::make_unique<ScopedCpuTensorHandle>(constTensor); + layer->m_QuantizedLstmParameters.m_InputToCellWeights = std::make_unique<ScopedTensorHandle>(constTensor); + layer->m_QuantizedLstmParameters.m_InputToForgetWeights = std::make_unique<ScopedTensorHandle>(constTensor); + layer->m_QuantizedLstmParameters.m_CellBias = std::make_unique<ScopedTensorHandle>(constTensor); + layer->m_QuantizedLstmParameters.m_ForgetGateBias = std::make_unique<ScopedTensorHandle>(constTensor); + layer->m_QuantizedLstmParameters.m_InputGateBias = std::make_unique<ScopedTensorHandle>(constTensor); + layer->m_QuantizedLstmParameters.m_OutputGateBias = std::make_unique<ScopedTensorHandle>(constTensor); + layer->m_QuantizedLstmParameters.m_RecurrentToForgetWeights = std::make_unique<ScopedTensorHandle>(constTensor); + layer->m_QuantizedLstmParameters.m_RecurrentToCellWeights = std::make_unique<ScopedTensorHandle>(constTensor); + layer->m_QuantizedLstmParameters.m_InputToOutputWeights = std::make_unique<ScopedTensorHandle>(constTensor); + layer->m_QuantizedLstmParameters.m_RecurrentToOutputWeights = std::make_unique<ScopedTensorHandle>(constTensor); + layer->m_QuantizedLstmParameters.m_RecurrentToInputWeights = std::make_unique<ScopedTensorHandle>(constTensor); + layer->m_QuantizedLstmParameters.m_InputToInputWeights = std::make_unique<ScopedTensorHandle>(constTensor); RunShapeInferenceTest<QuantizedLstmLayer>(layer, {{2, 20}, {2, 20}, {2, 20}}); } diff --git a/src/armnn/test/SubgraphViewTests.cpp b/src/armnn/test/SubgraphViewTests.cpp index 73ef8bea91..ecb876dc7a 100644 --- a/src/armnn/test/SubgraphViewTests.cpp +++ b/src/armnn/test/SubgraphViewTests.cpp @@ -10,7 +10,7 @@ #include <armnn/utility/NumericCast.hpp> -#include <backendsCommon/CpuTensorHandle.hpp> +#include <backendsCommon/TensorHandle.hpp> #include <fstream> #include <map> #include <queue> diff --git a/src/armnn/test/optimizations/AddBroadcastReshapeLayerTests.cpp b/src/armnn/test/optimizations/AddBroadcastReshapeLayerTests.cpp index 4523e70437..d0d728bfab 100644 --- a/src/armnn/test/optimizations/AddBroadcastReshapeLayerTests.cpp +++ b/src/armnn/test/optimizations/AddBroadcastReshapeLayerTests.cpp @@ -299,7 +299,7 @@ BOOST_AUTO_TEST_CASE(ReshapeParentConstLayerTest) uint8_t tensor[] = { 1, 1, 1, 1, 1 }; - constant->m_LayerOutput = std::make_unique<ScopedCpuTensorHandle>(ConstTensor(info1, &tensor)); + constant->m_LayerOutput = std::make_unique<ScopedTensorHandle>(ConstTensor(info1, &tensor)); input->GetOutputSlot().SetTensorInfo(info0); constant->GetOutputSlot().SetTensorInfo(info1); @@ -357,7 +357,7 @@ BOOST_AUTO_TEST_CASE(ReshapeParentConstAddLayerMultipleConnectionsTest) input->GetOutputSlot().SetTensorInfo(inputInfo); constant->GetOutputSlot().SetTensorInfo(constantTermInfo); float tensor[] = { 2.0f }; - constant->m_LayerOutput = std::make_unique<ScopedCpuTensorHandle>(ConstTensor(constantTermInfo, &tensor)); + constant->m_LayerOutput = std::make_unique<ScopedTensorHandle>(ConstTensor(constantTermInfo, &tensor)); add1->GetOutputSlot().SetTensorInfo(outputInfo); input->GetOutputSlot().Connect(add1->GetInputSlot(0)); diff --git a/src/armnn/test/optimizations/ConvertConstantsBFloatTests.cpp b/src/armnn/test/optimizations/ConvertConstantsBFloatTests.cpp index bb8e674b56..e4c1f2f413 100644 --- a/src/armnn/test/optimizations/ConvertConstantsBFloatTests.cpp +++ b/src/armnn/test/optimizations/ConvertConstantsBFloatTests.cpp @@ -38,7 +38,7 @@ BOOST_AUTO_TEST_CASE(ConvertConstantsFloatToBFloatTest) input->GetOutputSlot().SetTensorInfo(info); auto fc = graph.AddLayer<armnn::FullyConnectedLayer>(armnn::FullyConnectedDescriptor(), "fc"); - fc->m_Weight = std::make_unique<armnn::ScopedCpuTensorHandle>(weights); + fc->m_Weight = std::make_unique<armnn::ScopedTensorHandle>(weights); fc->GetOutputSlot().SetTensorInfo(info); auto output = graph.AddLayer<armnn::OutputLayer>(1, "output"); @@ -94,7 +94,7 @@ BOOST_AUTO_TEST_CASE(ConvertConstantsBFloatToFloatTest) input->GetOutputSlot().SetTensorInfo(info); auto fc = graph.AddLayer<armnn::FullyConnectedLayer>(armnn::FullyConnectedDescriptor(), "fc"); - fc->m_Weight = std::make_unique<armnn::ScopedCpuTensorHandle>(weights); + fc->m_Weight = std::make_unique<armnn::ScopedTensorHandle>(weights); fc->GetOutputSlot().SetTensorInfo(info); auto output = graph.AddLayer<armnn::OutputLayer>(1, "output"); diff --git a/src/armnn/test/optimizations/ConvertConstantsFloatToHalfTests.cpp b/src/armnn/test/optimizations/ConvertConstantsFloatToHalfTests.cpp index 12df462456..1dfe7f431c 100644 --- a/src/armnn/test/optimizations/ConvertConstantsFloatToHalfTests.cpp +++ b/src/armnn/test/optimizations/ConvertConstantsFloatToHalfTests.cpp @@ -31,7 +31,7 @@ BOOST_AUTO_TEST_CASE(ConvertConstantsFloatToHalfTest) input->GetOutputSlot().SetTensorInfo(info); auto fc = graph.AddLayer<armnn::FullyConnectedLayer>(armnn::FullyConnectedDescriptor(), "fc"); - fc->m_Weight = std::make_unique<armnn::ScopedCpuTensorHandle>(weights); + fc->m_Weight = std::make_unique<armnn::ScopedTensorHandle>(weights); fc->GetOutputSlot().SetTensorInfo(info); auto output = graph.AddLayer<armnn::OutputLayer>(1, "output"); diff --git a/src/armnn/test/optimizations/ConvertConstantsHalfToFloatTests.cpp b/src/armnn/test/optimizations/ConvertConstantsHalfToFloatTests.cpp index 7d7c6b2b0a..1ddf5262e8 100644 --- a/src/armnn/test/optimizations/ConvertConstantsHalfToFloatTests.cpp +++ b/src/armnn/test/optimizations/ConvertConstantsHalfToFloatTests.cpp @@ -31,7 +31,7 @@ BOOST_AUTO_TEST_CASE(ConvertConstantsHalfToFloatTest) input->GetOutputSlot().SetTensorInfo(info); auto fc = graph.AddLayer<armnn::FullyConnectedLayer>(armnn::FullyConnectedDescriptor(), "fc"); - fc->m_Weight = std::make_unique<armnn::ScopedCpuTensorHandle>(weights); + fc->m_Weight = std::make_unique<armnn::ScopedTensorHandle>(weights); fc->GetOutputSlot().SetTensorInfo(info); auto output = graph.AddLayer<armnn::OutputLayer>(1, "output"); diff --git a/src/armnn/test/optimizations/Fp32NetworkToBf16ConverterTests.cpp b/src/armnn/test/optimizations/Fp32NetworkToBf16ConverterTests.cpp index a65012eef4..f93fa77b0d 100644 --- a/src/armnn/test/optimizations/Fp32NetworkToBf16ConverterTests.cpp +++ b/src/armnn/test/optimizations/Fp32NetworkToBf16ConverterTests.cpp @@ -72,8 +72,8 @@ BOOST_AUTO_TEST_CASE(Fp32NetworkToBf16OptimizationConv2DTest) armnn::Convolution2dDescriptor descriptor; auto conv = graph.AddLayer<armnn::Convolution2dLayer>(descriptor, "conv2d"); - conv->m_Weight = std::make_unique<armnn::ScopedCpuTensorHandle>(weights); - conv->m_Bias = std::make_unique<armnn::ScopedCpuTensorHandle>(bias); + conv->m_Weight = std::make_unique<armnn::ScopedTensorHandle>(weights); + conv->m_Bias = std::make_unique<armnn::ScopedTensorHandle>(bias); conv->GetOutputSlot().SetTensorInfo(infoFP32); auto output = graph.AddLayer<armnn::OutputLayer>(1, "output"); @@ -142,8 +142,8 @@ BOOST_AUTO_TEST_CASE(Fp32NetworkToBf16OptimizationFullyConnectedTest) armnn::FullyConnectedDescriptor descriptor; auto fc = graph.AddLayer<armnn::FullyConnectedLayer>(descriptor, "fully"); - fc->m_Weight = std::make_unique<armnn::ScopedCpuTensorHandle>(weights); - fc->m_Bias = std::make_unique<armnn::ScopedCpuTensorHandle>(bias); + fc->m_Weight = std::make_unique<armnn::ScopedTensorHandle>(weights); + fc->m_Bias = std::make_unique<armnn::ScopedTensorHandle>(bias); fc->GetOutputSlot().SetTensorInfo(infoFP32); auto output = graph.AddLayer<armnn::OutputLayer>(1, "output"); |