IVGCVSW-5568 Revert "IVGCVSW-5563 Fix Crash on model with FullyConnected Sigmoid Activation"
* This reverts commit be25d94aefe53f221304b1f5f344913b708f808b.
* Add Unit Test: any receiver layer + any activation layer in float and QAsymmU8
* Tidy up fuse activation tests
Signed-off-by: Teresa Charlin <teresa.charlinreyes@arm.com>
Change-Id: Ie059d03b85cd17eaaafe5188bb173672a1fb9ae0
diff --git a/src/armnn/test/optimizations/FuseActivationTests.cpp b/src/armnn/test/optimizations/FuseActivationTests.cpp
index 0e85597..f28abff 100644
--- a/src/armnn/test/optimizations/FuseActivationTests.cpp
+++ b/src/armnn/test/optimizations/FuseActivationTests.cpp
@@ -19,10 +19,8 @@
BOOST_AUTO_TEST_SUITE(Optimizer)
-namespace
+namespace armnn
{
-const float g_qScale = 1.0f;
-const int32_t g_qOffset = 0;
template<typename T>
std::vector<T> GetVector(unsigned int size, float initial, float increment)
@@ -40,10 +38,10 @@
return vector;
}
-template<armnn::DataType ArmnnType, typename T = armnn::ResolveType<ArmnnType>>
+template<DataType ArmnnType, typename T = ResolveType<ArmnnType>>
struct Convolution2dTest
{
- using LayerType = armnn::Convolution2dLayer;
+ using LayerType = Convolution2dLayer;
static std::string GetReceiverLayerName() { return "Convolution2d"; };
static const bool isElementWise = false;
@@ -55,7 +53,9 @@
constexpr static const unsigned int outputSize = 36; // batchOut * heightOut * widthOut * channelOut
static IConnectableLayer* AddReceiverLayer(INetwork* network,
- const char* name)
+ const char* name,
+ float scale = 1.f,
+ int32_t offset = 0)
{
Convolution2dDescriptor descriptor;
descriptor.m_BiasEnabled = false;
@@ -67,8 +67,8 @@
11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22,
21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32,
31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42};
- std::vector<T> weightsVector = armnnUtils::QuantizedVector<T>(weightsData, g_qScale, g_qOffset);
- TensorInfo weightsInfo(GetWeightsShape(), ArmnnType, g_qScale, g_qOffset);
+ std::vector<T> weightsVector = armnnUtils::QuantizedVector<T>(weightsData, scale, offset);
+ TensorInfo weightsInfo(GetWeightsShape(), ArmnnType, scale, offset);
ConstTensor weights(weightsInfo, weightsVector);
Optional<ConstTensor> optionalBias;
@@ -76,11 +76,11 @@
}
};
-template<armnn::DataType ArmnnType, typename T = armnn::ResolveType<ArmnnType>>
-struct DepthwiseConvolution2dTest
+template<DataType ArmnnType, typename T = ResolveType<ArmnnType>>
+struct DWConvolution2dTest
{
public:
- using LayerType = armnn::DepthwiseConvolution2dLayer;
+ using LayerType = DepthwiseConvolution2dLayer;
static std::string GetReceiverLayerName() { return "DepthwiseConvolution2d"; };
static const bool isElementWise = false;
@@ -92,7 +92,9 @@
constexpr static const unsigned int outputSize = 108; //batchOut * heightOut * widthOut * channelOut;
static IConnectableLayer* AddReceiverLayer(INetwork* network,
- const char* name)
+ const char* name,
+ float scale = 1.f,
+ int32_t offset = 0)
{
DepthwiseConvolution2dDescriptor descriptor;
descriptor.m_BiasEnabled = false;
@@ -104,8 +106,8 @@
11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22,
21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32,
31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42};
- std::vector<T> weightsVector = armnnUtils::QuantizedVector<T>(weightsData, g_qScale, g_qOffset);
- TensorInfo weightsInfo(GetWeightsShape(), ArmnnType, g_qScale, g_qOffset);
+ std::vector<T> weightsVector = armnnUtils::QuantizedVector<T>(weightsData, scale, offset);
+ TensorInfo weightsInfo(GetWeightsShape(), ArmnnType, scale, offset);
ConstTensor weights(weightsInfo, weightsVector);
Optional<ConstTensor> optionalBias;
@@ -113,11 +115,11 @@
}
};
-template<armnn::DataType ArmnnType, typename T = armnn::ResolveType<ArmnnType>>
+template<DataType ArmnnType, typename T = ResolveType<ArmnnType>>
struct FullyConnectedTest
{
public:
- using LayerType = armnn::FullyConnectedLayer;
+ using LayerType = FullyConnectedLayer;
static std::string GetReceiverLayerName() { return "FullyConnected"; };
static const bool isElementWise = false;
@@ -129,7 +131,9 @@
constexpr static const unsigned int outputSize = 6; // batchOut * heightOut * widthOut * channelOut
static IConnectableLayer* AddReceiverLayer(INetwork* network,
- const char* name)
+ const char* name,
+ float scale = 1.f,
+ int32_t offset = 0)
{
FullyConnectedDescriptor descriptor;
descriptor.m_BiasEnabled = false;
@@ -137,8 +141,8 @@
std::vector<float> weightsData = { 1, 2, 3, 4, 5,
6, 7, 8, 9, 10,
11, 12, 13, 14, 15};
- std::vector<T> weightsVector = armnnUtils::QuantizedVector<T>(weightsData, g_qScale, g_qOffset);
- TensorInfo weightsInfo(GetWeightsShape(), ArmnnType, g_qScale, g_qOffset);
+ std::vector<T> weightsVector = armnnUtils::QuantizedVector<T>(weightsData, scale, offset);
+ TensorInfo weightsInfo(GetWeightsShape(), ArmnnType, scale, offset);
ConstTensor weights(weightsInfo, weightsVector);
Optional<ConstTensor> optionalBias;
@@ -146,11 +150,11 @@
}
};
-template<armnn::DataType ArmnnType, typename T = armnn::ResolveType<ArmnnType>>
+template<DataType ArmnnType, typename T = ResolveType<ArmnnType>>
struct BatchNormTest
{
public:
- using LayerType = armnn::BatchNormalizationLayer;
+ using LayerType = BatchNormalizationLayer;
static std::string GetReceiverLayerName() { return "BatchNorm"; };
static const bool isElementWise = false;
@@ -161,8 +165,13 @@
constexpr static const unsigned int outputSize = 48; // batchOut * heightOut * widthOut * channelOut
static IConnectableLayer* AddReceiverLayer(INetwork* network,
- const char* name)
+ const char* name,
+ float scale = 1.f,
+ int32_t offset = 0)
{
+ IgnoreUnused(scale);
+ IgnoreUnused(offset);
+
BatchNormalizationDescriptor descriptor;
descriptor.m_DataLayout = DataLayout::NHWC;
@@ -181,10 +190,10 @@
}
};
-template<armnn::DataType ArmnnType, typename T = armnn::ResolveType<ArmnnType>>
+template<DataType ArmnnType, typename T = ResolveType<ArmnnType>>
struct MultiplicationTest
{
- using LayerType = armnn::MultiplicationLayer;
+ using LayerType = MultiplicationLayer;
static std::string GetReceiverLayerName() { return "Multiplication"; };
static const bool isElementWise = true;
@@ -195,16 +204,21 @@
constexpr static const unsigned int outputSize = 48; // batchOut * heightOut * widthOut * channelOut
static IConnectableLayer* AddReceiverLayer(INetwork* network,
- const char* name)
+ const char* name,
+ float scale = 1.f,
+ int32_t offset = 0)
{
+ IgnoreUnused(scale);
+ IgnoreUnused(offset);
+
return network->AddMultiplicationLayer(name);
}
};
-template<armnn::DataType ArmnnType, typename T = armnn::ResolveType<ArmnnType>>
+template<DataType ArmnnType, typename T = ResolveType<ArmnnType>>
struct AdditionTest
{
- using LayerType = armnn::AdditionLayer;
+ using LayerType = AdditionLayer;
static std::string GetReceiverLayerName() { return "Addition"; };
static const bool isElementWise = true;
@@ -215,16 +229,21 @@
constexpr static const unsigned int outputSize = 48; // batchOut * heightOut * widthOut * channelOut
static IConnectableLayer* AddReceiverLayer(INetwork* network,
- const char* name)
+ const char* name,
+ float scale = 1.f,
+ int32_t offset = 0)
{
+ IgnoreUnused(scale);
+ IgnoreUnused(offset);
+
return network->AddAdditionLayer(name);
}
};
-template<armnn::DataType ArmnnType, typename T = armnn::ResolveType<ArmnnType>>
+template<DataType ArmnnType, typename T = ResolveType<ArmnnType>>
struct SubtractionTest
{
- using LayerType = armnn::SubtractionLayer;
+ using LayerType = SubtractionLayer;
static std::string GetReceiverLayerName() { return "Subtraction"; };
static const bool isElementWise = true;
@@ -235,16 +254,21 @@
constexpr static const unsigned int outputSize = 48; // batchOut * heightOut * widthOut * channelOut
static IConnectableLayer* AddReceiverLayer(INetwork* network,
- const char* name)
+ const char* name,
+ float scale = 1.f,
+ int32_t offset = 0)
{
+ IgnoreUnused(scale);
+ IgnoreUnused(offset);
+
return network->AddSubtractionLayer(name);
}
};
-template<armnn::DataType ArmnnType, typename T = armnn::ResolveType<ArmnnType>>
+template<DataType ArmnnType, typename T = ResolveType<ArmnnType>>
struct DivisionTest
{
- using LayerType = armnn::DivisionLayer;
+ using LayerType = DivisionLayer;
static std::string GetReceiverLayerName() { return "Division"; };
static const bool isElementWise = true;
@@ -255,17 +279,21 @@
constexpr static const unsigned int outputSize = 48; // batchOut * heightOut * widthOut * channelOut
static IConnectableLayer* AddReceiverLayer(INetwork* network,
- const char* name)
+ const char* name,
+ float scale = 1.f,
+ int32_t offset = 0)
{
+ IgnoreUnused(scale);
+ IgnoreUnused(offset);
+
return network->AddDivisionLayer(name);
}
};
-} // namespace
-
template<typename LayerTest,
- armnn::DataType ArmnnType>
-INetworkPtr CreatNetwork(ActivationDescriptor activationDescriptor, bool preventFusing)
+ DataType ArmnnType>
+INetworkPtr CreatNetwork(ActivationDescriptor activationDescriptor, bool preventFusing,
+ float scale, int32_t offset)
{
// Create a network
INetworkPtr network = INetwork::Create();
@@ -273,7 +301,9 @@
IConnectableLayer* inputLayer = network->AddInputLayer(0);
IConnectableLayer* receiverLayer = LayerTest::AddReceiverLayer(network.get(),
- "receiverLayer");
+ "receiverLayer",
+ scale,
+ offset);
IConnectableLayer* activationLayer = network->AddActivationLayer(activationDescriptor,
"activation");
@@ -282,8 +312,8 @@
IConnectableLayer* output2Layer = preventFusing?network->AddOutputLayer(1):nullptr;
// Define layers information
- TensorInfo inputInfo(LayerTest::GetInputShape(), ArmnnType, g_qScale, g_qOffset);
- TensorInfo outputInfo(LayerTest::GetOutputShape(), ArmnnType, g_qScale, g_qOffset);
+ TensorInfo inputInfo(LayerTest::GetInputShape(), ArmnnType, scale, offset);
+ TensorInfo outputInfo(LayerTest::GetOutputShape(), ArmnnType, scale, offset);
// Set layer information
inputLayer->GetOutputSlot(0).SetTensorInfo(inputInfo);
@@ -308,15 +338,15 @@
}
template<typename LayerTest,
- armnn::DataType ArmnnType,
+ DataType ArmnnType,
typename LayerType = typename LayerTest::LayerType,
- typename T = armnn::ResolveType<ArmnnType>>
-void FuseActivationIntoPreviousLayerTest(ActivationDescriptor activationDescriptor, float tolerance, armnn::Compute
-backendId)
+ typename T = ResolveType<ArmnnType>>
+void FuseActivationIntoPreviousLayerTest(ActivationDescriptor activationDescriptor, float tolerance, Compute backendId,
+ float scale = 1.f, int32_t offset=0)
{
// FIRST NETWORK: Fused
// Construct ArmNN network
- INetworkPtr networkFused = CreatNetwork<LayerTest, ArmnnType>(activationDescriptor, false);
+ INetworkPtr networkFused = CreatNetwork<LayerTest, ArmnnType>(activationDescriptor, false, scale, offset);
// Create ArmNN runtime
IRuntimePtr run = IRuntime::Create(IRuntime::CreationOptions()); // default options
@@ -326,7 +356,7 @@
Graph graphFused = PolymorphicDowncast<OptimizedNetwork*>(optNetFused.get())->GetGraph();
- auto checkFusedConv2d = [](const armnn::Layer* const layer)->bool {
+ auto checkFusedConv2d = [](const Layer* const layer)->bool {
return IsLayerOfType<LayerType>(layer) &&
(layer->GetNameStr() == "fused-activation-into-receiverLayer");
};
@@ -344,7 +374,7 @@
//Creates structures for inputs and outputs.
std::vector<float> data = GetVector<float>(LayerTest::inputSize, 1.0f, 0.1f);
- std::vector<T> inputDataFused = armnnUtils::QuantizedVector<T>(data, g_qScale, g_qOffset);
+ std::vector<T> inputDataFused = armnnUtils::QuantizedVector<T>(data, scale, offset);
std::vector<T> outputDataFused(LayerTest::outputSize);
InputTensors inputTensorsFused{
@@ -357,7 +387,7 @@
// SECOND NETWORK: NotFused
// Construct ArmNN network
- INetworkPtr networkNotFused = CreatNetwork<LayerTest, ArmnnType>(activationDescriptor, true);
+ INetworkPtr networkNotFused = CreatNetwork<LayerTest, ArmnnType>(activationDescriptor, true, scale, offset);
// Create ArmNN runtime
IRuntimePtr runNotFused = IRuntime::Create(IRuntime::CreationOptions()); // default options
@@ -370,18 +400,18 @@
BOOST_CHECK(5 == graphNotFused.GetNumLayers());
BOOST_TEST(CheckSequence(graphNotFused.cbegin(),
graphNotFused.cend(),
- &IsLayerOfType<armnn::InputLayer>,
+ &IsLayerOfType<InputLayer>,
&IsLayerOfType<LayerType>,
- &IsLayerOfType<armnn::ActivationLayer>,
- &IsLayerOfType<armnn::OutputLayer>,
- &IsLayerOfType<armnn::OutputLayer>));
+ &IsLayerOfType<ActivationLayer>,
+ &IsLayerOfType<OutputLayer>,
+ &IsLayerOfType<OutputLayer>));
// Load network into runtime
NetworkId networkIdentifierNotFused;
BOOST_TEST(runNotFused->LoadNetwork(networkIdentifierNotFused, std::move(optNetNotFused)) == Status::Success);
//Creates structures for inputs and outputs.
- std::vector<T> inputDataNotFused = armnnUtils::QuantizedVector<T>(data, g_qScale, g_qOffset);
+ std::vector<T> inputDataNotFused = armnnUtils::QuantizedVector<T>(data, scale, offset);
std::vector<T> outputDataNotFused(LayerTest::outputSize);
std::vector<T> outputData2NotFused(LayerTest::outputSize);
@@ -402,6 +432,58 @@
}
}
+template<typename LayerTest,
+ DataType ArmnnType,
+ typename LayerType = typename LayerTest::LayerType,
+ typename T = ResolveType<ArmnnType>>
+bool FuseActivationSimpleTest(ActivationDescriptor activationDescriptor, Compute backendId,
+ float scale = 1.f, int32_t offset = 0)
+{
+ bool success;
+ try
+ {
+ // Construct ArmNN network
+ INetworkPtr networkFused = CreatNetwork<LayerTest, ArmnnType>(activationDescriptor, false, scale, offset);
+
+ // Create ArmNN runtime
+ IRuntimePtr run = IRuntime::Create(IRuntime::CreationOptions()); // default options
+
+ // Optimise ArmNN network
+ IOptimizedNetworkPtr optNetFused = Optimize(*networkFused, {backendId}, run->GetDeviceSpec());
+
+ Graph graphFused = PolymorphicDowncast<OptimizedNetwork*>(optNetFused.get())->GetGraph();
+
+ // Load network into runtime
+ NetworkId networkIdentifier;
+ BOOST_TEST(run->LoadNetwork(networkIdentifier, std::move(optNetFused)) == Status::Success);
+
+ //Creates structures for inputs and outputs.
+ std::vector<float> data = GetVector<float>(LayerTest::inputSize, 1.0f, 0.1f);
+ std::vector<T> inputDataFused = armnnUtils::QuantizedVector<T>(data, scale, offset);
+ std::vector<T> outputDataFused(LayerTest::outputSize);
+
+ InputTensors inputTensorsFused{
+ {0, ConstTensor(run->GetInputTensorInfo(networkIdentifier, 0), inputDataFused.data())}};
+ OutputTensors outputTensorsFused{
+ {0, Tensor(run->GetOutputTensorInfo(networkIdentifier, 0), outputDataFused.data())}};
+
+ // Execute network
+ run->EnqueueWorkload(networkIdentifier, inputTensorsFused, outputTensorsFused);
+
+ success = true;
+ }
+ catch (const std::exception& e)
+ {
+ std::cerr << e.what() << std::endl;
+ success = false;
+ }
+
+ return success;
+}
+
+} // namespace armnn
+
+using namespace armnn;
#if defined(ARMCOMPUTENEON_ENABLED)
// ReLu fused into Receiver Layers Float32
BOOST_AUTO_TEST_CASE(FuseReLUIntoConvFloat32CpuAccTest)
@@ -410,15 +492,15 @@
activationDescriptor.m_Function = ActivationFunction::ReLu;
FuseActivationIntoPreviousLayerTest<Convolution2dTest<DataType::Float32>, DataType::Float32>
- (activationDescriptor, 0.0001f, armnn::Compute::CpuAcc);
+ (activationDescriptor, 0.0001f, Compute::CpuAcc);
}
BOOST_AUTO_TEST_CASE(FuseReLUIntoDWConvFloat32CpuAccTest)
{
ActivationDescriptor activationDescriptor;
activationDescriptor.m_Function = ActivationFunction::ReLu;
- FuseActivationIntoPreviousLayerTest<DepthwiseConvolution2dTest<DataType::Float32>, DataType::Float32>
- (activationDescriptor, 0.0001f, armnn::Compute::CpuAcc);
+ FuseActivationIntoPreviousLayerTest<DWConvolution2dTest<DataType::Float32>, DataType::Float32>
+ (activationDescriptor, 0.0001f, Compute::CpuAcc);
}
BOOST_AUTO_TEST_CASE(FuseReLUIntoFullyConnectedFloat32CpuAccTest)
{
@@ -426,7 +508,7 @@
activationDescriptor.m_Function = ActivationFunction::ReLu;
FuseActivationIntoPreviousLayerTest<FullyConnectedTest<DataType::Float32>, DataType::Float32>
- (activationDescriptor, 0.0001f, armnn::Compute::CpuAcc);
+ (activationDescriptor, 0.0001f, Compute::CpuAcc);
}
BOOST_AUTO_TEST_CASE(FuseReLUIntoBatchNormFloat32CpuAccTest)
{
@@ -434,7 +516,7 @@
activationDescriptor.m_Function = ActivationFunction::ReLu;
FuseActivationIntoPreviousLayerTest<BatchNormTest<DataType::Float32>, DataType::Float32>
- (activationDescriptor, 0.0001f, armnn::Compute::CpuAcc);
+ (activationDescriptor, 0.0001f, Compute::CpuAcc);
}
// BoundedReLu fused into Receiver Layers Float32
@@ -446,7 +528,7 @@
activationDescriptor.m_B = -1.0f;
FuseActivationIntoPreviousLayerTest<Convolution2dTest<DataType::Float32>, DataType::Float32>
- (activationDescriptor, 0.0001f, armnn::Compute::CpuAcc);
+ (activationDescriptor, 0.0001f, Compute::CpuAcc);
}
BOOST_AUTO_TEST_CASE(FuseBoundedReLUIntoDWConvFloat32CpuAccTest)
{
@@ -455,8 +537,8 @@
activationDescriptor.m_A = 1.0f;
activationDescriptor.m_B = -1.0f;
- FuseActivationIntoPreviousLayerTest < DepthwiseConvolution2dTest < DataType::Float32 > , DataType::Float32 >
- (activationDescriptor, 0.0001f, armnn::Compute::CpuAcc);
+ FuseActivationIntoPreviousLayerTest < DWConvolution2dTest < DataType::Float32 > , DataType::Float32 >
+ (activationDescriptor, 0.0001f, Compute::CpuAcc);
}
BOOST_AUTO_TEST_CASE(FuseBoundedReLUIntoFullyConnectedFloat32CpuAccTest)
{
@@ -466,7 +548,7 @@
activationDescriptor.m_B = -1.0f;
FuseActivationIntoPreviousLayerTest<FullyConnectedTest<DataType::Float32>, DataType::Float32>
- (activationDescriptor, 0.0001f, armnn::Compute::CpuAcc);
+ (activationDescriptor, 0.0001f, Compute::CpuAcc);
}
BOOST_AUTO_TEST_CASE(FuseBoundedReLUIntoBatchNormFloat32CpuAccTest)
{
@@ -476,7 +558,7 @@
activationDescriptor.m_B = -1.0f;
FuseActivationIntoPreviousLayerTest<BatchNormTest<DataType::Float32>, DataType::Float32>
- (activationDescriptor, 0.0001f, armnn::Compute::CpuAcc);
+ (activationDescriptor, 0.0001f, Compute::CpuAcc);
}
// ReLU fused into Receiver Layers QAsymmU8
@@ -486,15 +568,15 @@
activationDescriptor.m_Function = ActivationFunction::ReLu;
FuseActivationIntoPreviousLayerTest<Convolution2dTest<DataType::QAsymmU8>, DataType::QAsymmU8>
- (activationDescriptor, 0.0001f, armnn::Compute::CpuAcc);
+ (activationDescriptor, 0.0001f, Compute::CpuAcc);
}
BOOST_AUTO_TEST_CASE(FuseReLUIntoDWConvQAsymmU8CpuAccTest)
{
ActivationDescriptor activationDescriptor;
activationDescriptor.m_Function = ActivationFunction::ReLu;
- FuseActivationIntoPreviousLayerTest<DepthwiseConvolution2dTest<DataType::QAsymmU8>, DataType::QAsymmU8>
- (activationDescriptor, 0.0001f, armnn::Compute::CpuAcc);
+ FuseActivationIntoPreviousLayerTest<DWConvolution2dTest<DataType::QAsymmU8>, DataType::QAsymmU8>
+ (activationDescriptor, 0.0001f, Compute::CpuAcc);
}
BOOST_AUTO_TEST_CASE(FuseReLUIntoFullyConnectedQAsymmU8CpuAccTest)
{
@@ -502,7 +584,7 @@
activationDescriptor.m_Function = ActivationFunction::ReLu;
FuseActivationIntoPreviousLayerTest<FullyConnectedTest<DataType::QAsymmU8>, DataType::QAsymmU8>
- (activationDescriptor, 0.0001f, armnn::Compute::CpuAcc);
+ (activationDescriptor, 0.0001f, Compute::CpuAcc);
}
// HardSwish fused into Receiver Layers Float32
@@ -512,7 +594,7 @@
activationDescriptor.m_Function = ActivationFunction::HardSwish;
FuseActivationIntoPreviousLayerTest<Convolution2dTest<DataType::Float32>, DataType::Float32>
- (activationDescriptor, 0.0001f, armnn::Compute::CpuAcc);
+ (activationDescriptor, 0.0001f, Compute::CpuAcc);
}
// TanH fused into Receiver Layers Float32
@@ -522,7 +604,91 @@
activationDescriptor.m_Function = ActivationFunction::TanH;
FuseActivationIntoPreviousLayerTest<Convolution2dTest<DataType::Float32>, DataType::Float32>
- (activationDescriptor, 0.0001f, armnn::Compute::CpuAcc);
+ (activationDescriptor, 0.0001f, Compute::CpuAcc);
+}
+
+// Test that all receiver layers follow by all activation layers work, either fused or not fused
+BOOST_AUTO_TEST_CASE(LayerFollowedByActivationFloat32CpuAccTest)
+{
+ ActivationDescriptor activationDescriptor;
+ for (int i = 0; i != 12; ++i)
+ {
+ activationDescriptor.m_Function = static_cast<ActivationFunction>(i);
+ activationDescriptor.m_A = 1.0f;
+ activationDescriptor.m_B = -1.0f;
+ BOOST_CHECK_MESSAGE((FuseActivationSimpleTest<Convolution2dTest<DataType::Float32>, DataType::Float32>
+ (activationDescriptor, Compute::CpuAcc)), "Convolution + Activation function " << i);
+ BOOST_CHECK_MESSAGE((FuseActivationSimpleTest<DWConvolution2dTest<DataType::Float32>, DataType::Float32>
+ (activationDescriptor, Compute::CpuAcc)), "DepthwiseConvolution + Activation function " << i);
+ BOOST_CHECK_MESSAGE((FuseActivationSimpleTest<FullyConnectedTest<DataType::Float32>, DataType::Float32>
+ (activationDescriptor, Compute::CpuAcc)), "FullyConnected + Activation function " << i);
+ BOOST_CHECK_MESSAGE((FuseActivationSimpleTest<BatchNormTest<DataType::Float32>, DataType::Float32>
+ (activationDescriptor, Compute::CpuAcc)), "BatchNorm + Activation function " << i);
+ }
+}
+BOOST_AUTO_TEST_CASE(LayerFollowedByActivationFloat16CpuAccTest)
+{
+ ActivationDescriptor activationDescriptor;
+ for (int i = 0; i != 12; ++i)
+ {
+ activationDescriptor.m_Function = static_cast<ActivationFunction>(i);
+ activationDescriptor.m_A = 1.0f;
+ activationDescriptor.m_B = -1.0f;
+ BOOST_CHECK_MESSAGE((FuseActivationSimpleTest<Convolution2dTest<DataType::Float16>, DataType::Float16>
+ (activationDescriptor, Compute::CpuAcc)), "Convolution + Activation function " << i);
+ BOOST_CHECK_MESSAGE((FuseActivationSimpleTest<DWConvolution2dTest<DataType::Float16>, DataType::Float16>
+ (activationDescriptor, Compute::CpuAcc)), "DepthwiseConvolution + Activation function " << i);
+ BOOST_CHECK_MESSAGE((FuseActivationSimpleTest<FullyConnectedTest<DataType::Float16>, DataType::Float16>
+ (activationDescriptor, Compute::CpuAcc)), "FullyConnected + Activation function " << i);
+ BOOST_CHECK_MESSAGE((FuseActivationSimpleTest<BatchNormTest<DataType::Float16>, DataType::Float16>
+ (activationDescriptor, Compute::CpuAcc)), "BatchNorm + Activation function " << i);
+ }
+}
+BOOST_AUTO_TEST_CASE(LayerFollowedByActivationQAsymmU8CpuAccTest)
+{
+ ActivationDescriptor activationDescriptor;
+
+ activationDescriptor.m_Function = ActivationFunction::Sigmoid;
+ BOOST_CHECK_MESSAGE((FuseActivationSimpleTest<Convolution2dTest<DataType::QAsymmU8>, DataType::QAsymmU8>
+ (activationDescriptor, Compute::CpuAcc, 1.f / 256.f, 0)), "Convolution + Activation function " <<
+ static_cast<int>(activationDescriptor.m_Function));
+ BOOST_CHECK_MESSAGE((FuseActivationSimpleTest<FullyConnectedTest<DataType::QAsymmU8>, DataType::QAsymmU8>
+ (activationDescriptor, Compute::CpuAcc, 1.f / 256.f, 0)), "FullyConnected + Activation function " <<
+ static_cast<int>(activationDescriptor.m_Function));
+
+ activationDescriptor.m_Function = ActivationFunction::TanH;
+ BOOST_CHECK_MESSAGE((FuseActivationSimpleTest<Convolution2dTest<DataType::QAsymmU8>, DataType::QAsymmU8>
+ (activationDescriptor, Compute::CpuAcc, 1.f / 128.f, 128)), "Convolution + Activation function " <<
+ static_cast<int>(activationDescriptor.m_Function));
+ BOOST_CHECK_MESSAGE((FuseActivationSimpleTest<FullyConnectedTest<DataType::QAsymmU8>, DataType::QAsymmU8>
+ (activationDescriptor, Compute::CpuAcc, 1.f / 128.f, 128)), "FullyConnected + Activation function " <<
+ static_cast<int>(activationDescriptor.m_Function));
+
+ activationDescriptor.m_Function = ActivationFunction::ReLu;
+ BOOST_CHECK_MESSAGE((FuseActivationSimpleTest<Convolution2dTest<DataType::QAsymmU8>, DataType::QAsymmU8>
+ (activationDescriptor, Compute::CpuAcc)), "Convolution + Activation function " <<
+ static_cast<int>(activationDescriptor.m_Function));
+ BOOST_CHECK_MESSAGE((FuseActivationSimpleTest<FullyConnectedTest<DataType::QAsymmU8>, DataType::QAsymmU8>
+ (activationDescriptor, Compute::CpuAcc)), "FullyConnected + Activation function " <<
+ static_cast<int>(activationDescriptor.m_Function));
+
+ activationDescriptor.m_Function = ActivationFunction::BoundedReLu;
+ activationDescriptor.m_A = 1.0f;
+ activationDescriptor.m_B = -1.0f;
+ BOOST_CHECK_MESSAGE((FuseActivationSimpleTest<Convolution2dTest<DataType::QAsymmU8>, DataType::QAsymmU8>
+ (activationDescriptor, Compute::CpuAcc)), "Convolution + Activation function " <<
+ static_cast<int>(activationDescriptor.m_Function));
+ BOOST_CHECK_MESSAGE((FuseActivationSimpleTest<FullyConnectedTest<DataType::QAsymmU8>, DataType::QAsymmU8>
+ (activationDescriptor, Compute::CpuAcc)), "FullyConnected + Activation function " <<
+ static_cast<int>(activationDescriptor.m_Function));
+
+ activationDescriptor.m_Function = ActivationFunction::HardSwish;
+ BOOST_CHECK_MESSAGE((FuseActivationSimpleTest<Convolution2dTest<DataType::QAsymmU8>, DataType::QAsymmU8>
+ (activationDescriptor, Compute::CpuAcc)), "Convolution + Activation function " <<
+ static_cast<int>(activationDescriptor.m_Function));
+ BOOST_CHECK_MESSAGE((FuseActivationSimpleTest<FullyConnectedTest<DataType::QAsymmU8>, DataType::QAsymmU8>
+ (activationDescriptor, Compute::CpuAcc)), "FullyConnected + Activation function " <<
+ static_cast<int>(activationDescriptor.m_Function));
}
#endif
@@ -534,15 +700,15 @@
activationDescriptor.m_Function = ActivationFunction::ReLu;
FuseActivationIntoPreviousLayerTest<Convolution2dTest<DataType::Float32>, DataType::Float32>
- (activationDescriptor, 0.0001f, armnn::Compute::GpuAcc);
+ (activationDescriptor, 0.0001f, Compute::GpuAcc);
}
BOOST_AUTO_TEST_CASE(FuseReLUIntoDWConvFloat32GpuAccTest)
{
ActivationDescriptor activationDescriptor;
activationDescriptor.m_Function = ActivationFunction::ReLu;
- FuseActivationIntoPreviousLayerTest<DepthwiseConvolution2dTest<DataType::Float32>, DataType::Float32>
- (activationDescriptor, 0.0001f, armnn::Compute::GpuAcc);
+ FuseActivationIntoPreviousLayerTest<DWConvolution2dTest<DataType::Float32>, DataType::Float32>
+ (activationDescriptor, 0.0001f, Compute::GpuAcc);
}
BOOST_AUTO_TEST_CASE(FuseReLUIntoFullyConnectedFloat32GpuAccTest)
{
@@ -550,7 +716,7 @@
activationDescriptor.m_Function = ActivationFunction::ReLu;
FuseActivationIntoPreviousLayerTest<FullyConnectedTest<DataType::Float32>, DataType::Float32>
- (activationDescriptor, 0.0001f, armnn::Compute::GpuAcc);
+ (activationDescriptor, 0.0001f, Compute::GpuAcc);
}
BOOST_AUTO_TEST_CASE(FuseReLUIntoBatchNormFloat32GpuAccTest)
{
@@ -558,7 +724,7 @@
activationDescriptor.m_Function = ActivationFunction::ReLu;
FuseActivationIntoPreviousLayerTest<BatchNormTest<DataType::Float32>, DataType::Float32>
- (activationDescriptor, 0.0001f, armnn::Compute::GpuAcc);
+ (activationDescriptor, 0.0001f, Compute::GpuAcc);
}
BOOST_AUTO_TEST_CASE(FuseReLUIntoMulFloat32GpuAccTest)
{
@@ -566,7 +732,7 @@
activationDescriptor.m_Function = ActivationFunction::ReLu;
FuseActivationIntoPreviousLayerTest<MultiplicationTest<DataType::Float32>, DataType::Float32>
- (activationDescriptor, 0.0001f, armnn::Compute::GpuAcc);
+ (activationDescriptor, 0.0001f, Compute::GpuAcc);
}
BOOST_AUTO_TEST_CASE(FuseReLUIntoAddFloat32GpuAccTest)
{
@@ -574,7 +740,7 @@
activationDescriptor.m_Function = ActivationFunction::ReLu;
FuseActivationIntoPreviousLayerTest<AdditionTest<DataType::Float32>, DataType::Float32>
- (activationDescriptor, 0.0001f, armnn::Compute::GpuAcc);
+ (activationDescriptor, 0.0001f, Compute::GpuAcc);
}
BOOST_AUTO_TEST_CASE(FuseReLUIntoSubFloat32GpuAccTest)
{
@@ -582,7 +748,7 @@
activationDescriptor.m_Function = ActivationFunction::ReLu;
FuseActivationIntoPreviousLayerTest<SubtractionTest<DataType::Float32>, DataType::Float32>
- (activationDescriptor, 0.0001f, armnn::Compute::GpuAcc);
+ (activationDescriptor, 0.0001f, Compute::GpuAcc);
}
BOOST_AUTO_TEST_CASE(FuseReLUIntoDivFloat32GpuAccTest)
{
@@ -590,7 +756,7 @@
activationDescriptor.m_Function = ActivationFunction::ReLu;
FuseActivationIntoPreviousLayerTest<DivisionTest<DataType::Float32>, DataType::Float32>
- (activationDescriptor, 0.0001f, armnn::Compute::GpuAcc);
+ (activationDescriptor, 0.0001f, Compute::GpuAcc);
}
// BoundedReLu fused into Receiver Layers Float32
@@ -602,7 +768,7 @@
activationDescriptor.m_B = -1.0f;
FuseActivationIntoPreviousLayerTest<Convolution2dTest<DataType::Float32>, DataType::Float32>
- (activationDescriptor, 0.0001f, armnn::Compute::GpuAcc);
+ (activationDescriptor, 0.0001f, Compute::GpuAcc);
}
BOOST_AUTO_TEST_CASE(FuseBoundedReLUIntoDWConvFloat32GpuAccTest)
{
@@ -611,8 +777,8 @@
activationDescriptor.m_A = 1.0f;
activationDescriptor.m_B = -1.0f;
- FuseActivationIntoPreviousLayerTest<DepthwiseConvolution2dTest<DataType::Float32>, DataType::Float32>
- (activationDescriptor, 0.0001f, armnn::Compute::GpuAcc);
+ FuseActivationIntoPreviousLayerTest<DWConvolution2dTest<DataType::Float32>, DataType::Float32>
+ (activationDescriptor, 0.0001f, Compute::GpuAcc);
}
BOOST_AUTO_TEST_CASE(FuseBoundedReLUIntoFullyConnectedFloat32GpuAccTest)
{
@@ -622,7 +788,7 @@
activationDescriptor.m_B = -1.0f;
FuseActivationIntoPreviousLayerTest<FullyConnectedTest<DataType::Float32>, DataType::Float32>
- (activationDescriptor, 0.0001f, armnn::Compute::GpuAcc);
+ (activationDescriptor, 0.0001f, Compute::GpuAcc);
}
BOOST_AUTO_TEST_CASE(FuseBoundedReLUIntoBatchNormFloat32GpuAccTest)
{
@@ -632,7 +798,7 @@
activationDescriptor.m_B = -1.0f;
FuseActivationIntoPreviousLayerTest<BatchNormTest<DataType::Float32>, DataType::Float32>
- (activationDescriptor, 0.0001f, armnn::Compute::GpuAcc);
+ (activationDescriptor, 0.0001f, Compute::GpuAcc);
}
BOOST_AUTO_TEST_CASE(FuseBoundedReLUIntoMulFloat32GpuAccTest)
{
@@ -642,7 +808,7 @@
activationDescriptor.m_B = -1.0f;
FuseActivationIntoPreviousLayerTest<MultiplicationTest<DataType::Float32>, DataType::Float32>
- (activationDescriptor, 0.0001f, armnn::Compute::GpuAcc);
+ (activationDescriptor, 0.0001f, Compute::GpuAcc);
}
BOOST_AUTO_TEST_CASE(FuseBoundedReLUIntoAddFloat32GpuAccTest)
{
@@ -652,7 +818,7 @@
activationDescriptor.m_B = -1.0f;
FuseActivationIntoPreviousLayerTest<AdditionTest<DataType::Float32>, DataType::Float32>
- (activationDescriptor, 0.0001f, armnn::Compute::GpuAcc);
+ (activationDescriptor, 0.0001f, Compute::GpuAcc);
}
BOOST_AUTO_TEST_CASE(FuseBoundedReLUIntoSubFloat32GpuAccTest)
{
@@ -662,7 +828,7 @@
activationDescriptor.m_B = -1.0f;
FuseActivationIntoPreviousLayerTest<SubtractionTest<DataType::Float32>, DataType::Float32>
- (activationDescriptor, 0.0001f, armnn::Compute::GpuAcc);
+ (activationDescriptor, 0.0001f, Compute::GpuAcc);
}
BOOST_AUTO_TEST_CASE(FuseBoundedReLUIntoDivFloat32GpuAccTest)
{
@@ -672,7 +838,7 @@
activationDescriptor.m_B = -1.0f;
FuseActivationIntoPreviousLayerTest<DivisionTest<DataType::Float32>, DataType::Float32>
- (activationDescriptor, 0.0001f, armnn::Compute::GpuAcc);
+ (activationDescriptor, 0.0001f, Compute::GpuAcc);
}
// ReLU fused into Receiver Layers QAsymmU8
@@ -682,15 +848,15 @@
activationDescriptor.m_Function = ActivationFunction::ReLu;
FuseActivationIntoPreviousLayerTest<Convolution2dTest<DataType::QAsymmU8>, DataType::QAsymmU8>
- (activationDescriptor, 0.0001f, armnn::Compute::GpuAcc);
+ (activationDescriptor, 0.0001f, Compute::GpuAcc);
}
BOOST_AUTO_TEST_CASE(FuseReLUQIntoDWConvAsymmU8GpuAccTest)
{
ActivationDescriptor activationDescriptor;
activationDescriptor.m_Function = ActivationFunction::ReLu;
- FuseActivationIntoPreviousLayerTest<DepthwiseConvolution2dTest<DataType::QAsymmU8>, DataType::QAsymmU8>
- (activationDescriptor, 0.0001f, armnn::Compute::GpuAcc);
+ FuseActivationIntoPreviousLayerTest<DWConvolution2dTest<DataType::QAsymmU8>, DataType::QAsymmU8>
+ (activationDescriptor, 0.0001f, Compute::GpuAcc);
}
BOOST_AUTO_TEST_CASE(FuseReLUQIntoFullyConnectedAsymmU8GpuAccTest)
{
@@ -698,7 +864,7 @@
activationDescriptor.m_Function = ActivationFunction::ReLu;
FuseActivationIntoPreviousLayerTest<FullyConnectedTest<DataType::QAsymmU8>, DataType::QAsymmU8>
- (activationDescriptor, 0.0001f, armnn::Compute::GpuAcc);
+ (activationDescriptor, 0.0001f, Compute::GpuAcc);
}
// HardSwish fused into Receiver Layers Float32
@@ -708,7 +874,7 @@
activationDescriptor.m_Function = ActivationFunction::HardSwish;
FuseActivationIntoPreviousLayerTest<Convolution2dTest<DataType::Float32>, DataType::Float32>
- (activationDescriptor, 0.0001f, armnn::Compute::GpuAcc);
+ (activationDescriptor, 0.0001f, Compute::GpuAcc);
}
BOOST_AUTO_TEST_CASE(FuseHardSwishIntoMulFloat32GpuAccTest)
{
@@ -716,7 +882,7 @@
activationDescriptor.m_Function = ActivationFunction::HardSwish;
FuseActivationIntoPreviousLayerTest<MultiplicationTest<DataType::Float32>, DataType::Float32>
- (activationDescriptor, 0.0001f, armnn::Compute::GpuAcc);
+ (activationDescriptor, 0.0001f, Compute::GpuAcc);
}
BOOST_AUTO_TEST_CASE(FuseHardSwishIntoAddFloat32GpuAccTest)
{
@@ -724,7 +890,7 @@
activationDescriptor.m_Function = ActivationFunction::HardSwish;
FuseActivationIntoPreviousLayerTest<AdditionTest<DataType::Float32>, DataType::Float32>
- (activationDescriptor, 0.0001f, armnn::Compute::GpuAcc);
+ (activationDescriptor, 0.0001f, Compute::GpuAcc);
}
BOOST_AUTO_TEST_CASE(FuseHardSwishIntoSubFloat32GpuAccTest)
{
@@ -732,7 +898,7 @@
activationDescriptor.m_Function = ActivationFunction::HardSwish;
FuseActivationIntoPreviousLayerTest<SubtractionTest<DataType::Float32>, DataType::Float32>
- (activationDescriptor, 0.0001f, armnn::Compute::GpuAcc);
+ (activationDescriptor, 0.0001f, Compute::GpuAcc);
}
BOOST_AUTO_TEST_CASE(FuseHardSwishIntoDivFloat32GpuAccTest)
{
@@ -740,7 +906,7 @@
activationDescriptor.m_Function = ActivationFunction::HardSwish;
FuseActivationIntoPreviousLayerTest<DivisionTest<DataType::Float32>, DataType::Float32>
- (activationDescriptor, 0.0001f, armnn::Compute::GpuAcc);
+ (activationDescriptor, 0.0001f, Compute::GpuAcc);
}
// TanH fused into Receiver Layers Float32
@@ -750,7 +916,7 @@
activationDescriptor.m_Function = ActivationFunction::TanH;
FuseActivationIntoPreviousLayerTest<Convolution2dTest<DataType::Float32>, DataType::Float32>
- (activationDescriptor, 0.0001f, armnn::Compute::GpuAcc);
+ (activationDescriptor, 0.0001f, Compute::GpuAcc);
}
BOOST_AUTO_TEST_CASE(FuseTanHIntoMulFloat32GpuAccTest)
{
@@ -758,7 +924,7 @@
activationDescriptor.m_Function = ActivationFunction::TanH;
FuseActivationIntoPreviousLayerTest<MultiplicationTest<DataType::Float32>, DataType::Float32>
- (activationDescriptor, 0.0001f, armnn::Compute::GpuAcc);
+ (activationDescriptor, 0.0001f, Compute::GpuAcc);
}
BOOST_AUTO_TEST_CASE(FuseTanHIntoAddFloat32GpuAccTest)
{
@@ -766,7 +932,7 @@
activationDescriptor.m_Function = ActivationFunction::TanH;
FuseActivationIntoPreviousLayerTest<AdditionTest<DataType::Float32>, DataType::Float32>
- (activationDescriptor, 0.0001f, armnn::Compute::GpuAcc);
+ (activationDescriptor, 0.0001f, Compute::GpuAcc);
}
BOOST_AUTO_TEST_CASE(FuseTanHIntoSubFloat32GpuAccTest)
{
@@ -774,7 +940,7 @@
activationDescriptor.m_Function = ActivationFunction::TanH;
FuseActivationIntoPreviousLayerTest<SubtractionTest<DataType::Float32>, DataType::Float32>
- (activationDescriptor, 0.0001f, armnn::Compute::GpuAcc);
+ (activationDescriptor, 0.0001f, Compute::GpuAcc);
}
BOOST_AUTO_TEST_CASE(FuseTanHIntoDivFloat32GpuAccTest)
{
@@ -782,7 +948,113 @@
activationDescriptor.m_Function = ActivationFunction::TanH;
FuseActivationIntoPreviousLayerTest<DivisionTest<DataType::Float32>, DataType::Float32>
- (activationDescriptor, 0.0001f, armnn::Compute::GpuAcc);
+ (activationDescriptor, 0.0001f, Compute::GpuAcc);
+}
+
+// Test that all receiver layers follow by all activation layers work, either fused or not fused
+BOOST_AUTO_TEST_CASE(LayerFollowedByActivationFloat32GpuAccTest)
+{
+ ActivationDescriptor activationDescriptor;
+ for (int i = 0; i != 12; ++i)
+ {
+ activationDescriptor.m_Function = static_cast<ActivationFunction>(i);
+ activationDescriptor.m_A = 1.0f;
+ activationDescriptor.m_B = -1.0f;
+ if (activationDescriptor.m_Function != ActivationFunction::Elu)
+ {
+ BOOST_CHECK_MESSAGE((FuseActivationSimpleTest<Convolution2dTest<DataType::Float32>, DataType::Float32>
+ (activationDescriptor, Compute::GpuAcc)), "Convolution + Activation function " << i);
+ BOOST_CHECK_MESSAGE((FuseActivationSimpleTest<DWConvolution2dTest<DataType::Float32>, DataType::Float32>
+ (activationDescriptor, Compute::GpuAcc)), "DepthwiseConvolution + Activation function " << i);
+ BOOST_CHECK_MESSAGE((FuseActivationSimpleTest<FullyConnectedTest<DataType::Float32>, DataType::Float32>
+ (activationDescriptor, Compute::GpuAcc)), "FullyConnected + Activation function " << i);
+ BOOST_CHECK_MESSAGE((FuseActivationSimpleTest<BatchNormTest<DataType::Float32>, DataType::Float32>
+ (activationDescriptor, Compute::GpuAcc)), "BatchNorm + Activation function " << i);
+ BOOST_CHECK_MESSAGE((FuseActivationSimpleTest<MultiplicationTest<DataType::Float32>, DataType::Float32>
+ (activationDescriptor, Compute::GpuAcc)), "Multiplication + Activation function " << i);
+ BOOST_CHECK_MESSAGE((FuseActivationSimpleTest<AdditionTest<DataType::Float32>, DataType::Float32>
+ (activationDescriptor, Compute::GpuAcc)), "Addition + Activation function " << i);
+ BOOST_CHECK_MESSAGE((FuseActivationSimpleTest<SubtractionTest<DataType::Float32>, DataType::Float32>
+ (activationDescriptor, Compute::GpuAcc)), "Subtraction + Activation function " << i);
+ BOOST_CHECK_MESSAGE((FuseActivationSimpleTest<DivisionTest<DataType::Float32>, DataType::Float32>
+ (activationDescriptor, Compute::GpuAcc)), "Division + Activation function " << i);
+ }
+ }
+}
+BOOST_AUTO_TEST_CASE(LayerFollowedByActivationFloat16GpuAccTest)
+{
+ ActivationDescriptor activationDescriptor;
+ for (int i = 0; i != 12; ++i)
+ {
+ activationDescriptor.m_Function = static_cast<ActivationFunction>(i);
+ activationDescriptor.m_A = 1.0f;
+ activationDescriptor.m_B = -1.0f;
+ if (activationDescriptor.m_Function != ActivationFunction::Elu)
+ {
+ BOOST_CHECK_MESSAGE((FuseActivationSimpleTest<Convolution2dTest<DataType::Float16>, DataType::Float16>
+ (activationDescriptor, Compute::GpuAcc)), "Convolution + Activation function " << i);
+ BOOST_CHECK_MESSAGE((FuseActivationSimpleTest<DWConvolution2dTest<DataType::Float16>, DataType::Float16>
+ (activationDescriptor, Compute::GpuAcc)), "Depthwise + Activation function " << i);
+ BOOST_CHECK_MESSAGE((FuseActivationSimpleTest<FullyConnectedTest<DataType::Float16>, DataType::Float16>
+ (activationDescriptor, Compute::GpuAcc)), "FullyConnected + Activation function " << i);
+ BOOST_CHECK_MESSAGE((FuseActivationSimpleTest<BatchNormTest<DataType::Float16>, DataType::Float16>
+ (activationDescriptor, Compute::GpuAcc)), "BatchNorm + Activation function " << i);
+ BOOST_CHECK_MESSAGE((FuseActivationSimpleTest<MultiplicationTest<DataType::Float16>, DataType::Float16>
+ (activationDescriptor, Compute::GpuAcc)), "Multiplication + Activation function " << i);
+ BOOST_CHECK_MESSAGE((FuseActivationSimpleTest<AdditionTest<DataType::Float16>, DataType::Float16>
+ (activationDescriptor, Compute::GpuAcc)), "Addition + Activation function " << i);
+ BOOST_CHECK_MESSAGE((FuseActivationSimpleTest<SubtractionTest<DataType::Float16>, DataType::Float16>
+ (activationDescriptor, Compute::GpuAcc)), "Subtraction + Activation function " << i);
+ BOOST_CHECK_MESSAGE((FuseActivationSimpleTest<DivisionTest<DataType::Float16>, DataType::Float16>
+ (activationDescriptor, Compute::GpuAcc)), "Division + Activation function " << i);
+ }
+ }
+}
+BOOST_AUTO_TEST_CASE(LayerFollowedByActivationQAsymmU8GpuAccTest)
+{
+ ActivationDescriptor activationDescriptor;
+
+ activationDescriptor.m_Function = ActivationFunction::Sigmoid;
+ BOOST_CHECK_MESSAGE((FuseActivationSimpleTest<Convolution2dTest<DataType::QAsymmU8>, DataType::QAsymmU8>
+ (activationDescriptor, Compute::GpuAcc, 1.f / 256.f, 0)), "Convolution + Activation function " <<
+ static_cast<int>(activationDescriptor.m_Function));
+ BOOST_CHECK_MESSAGE((FuseActivationSimpleTest<FullyConnectedTest<DataType::QAsymmU8>, DataType::QAsymmU8>
+ (activationDescriptor, Compute::GpuAcc, 1.f / 256.f, 0)), "FullyConnected + Activation function " <<
+ static_cast<int>(activationDescriptor.m_Function));
+
+ activationDescriptor.m_Function = ActivationFunction::TanH;
+ BOOST_CHECK_MESSAGE((FuseActivationSimpleTest<Convolution2dTest<DataType::QAsymmU8>, DataType::QAsymmU8>
+ (activationDescriptor, Compute::GpuAcc, 1.f / 128.f, 128)), "Convolution + Activation function " <<
+ static_cast<int>(activationDescriptor.m_Function));
+ BOOST_CHECK_MESSAGE((FuseActivationSimpleTest<FullyConnectedTest<DataType::QAsymmU8>, DataType::QAsymmU8>
+ (activationDescriptor, Compute::GpuAcc, 1.f / 128.f, 128)), "FullyConnected + Activation function " <<
+ static_cast<int>(activationDescriptor.m_Function));
+
+ activationDescriptor.m_Function = ActivationFunction::ReLu;
+ BOOST_CHECK_MESSAGE((FuseActivationSimpleTest<Convolution2dTest<DataType::QAsymmU8>, DataType::QAsymmU8>
+ (activationDescriptor, Compute::GpuAcc)), "Convolution + Activation function " <<
+ static_cast<int>(activationDescriptor.m_Function));
+ BOOST_CHECK_MESSAGE((FuseActivationSimpleTest<FullyConnectedTest<DataType::QAsymmU8>, DataType::QAsymmU8>
+ (activationDescriptor, Compute::GpuAcc)), "FullyConnected + Activation function " <<
+ static_cast<int>(activationDescriptor.m_Function));
+
+ activationDescriptor.m_Function = ActivationFunction::BoundedReLu;
+ activationDescriptor.m_A = 1.0f;
+ activationDescriptor.m_B = -1.0f;
+ BOOST_CHECK_MESSAGE((FuseActivationSimpleTest<Convolution2dTest<DataType::QAsymmU8>, DataType::QAsymmU8>
+ (activationDescriptor, Compute::GpuAcc)), "Convolution + Activation function " <<
+ static_cast<int>(activationDescriptor.m_Function));
+ BOOST_CHECK_MESSAGE((FuseActivationSimpleTest<FullyConnectedTest<DataType::QAsymmU8>, DataType::QAsymmU8>
+ (activationDescriptor, Compute::GpuAcc)), "FullyConnected + Activation function " <<
+ static_cast<int>(activationDescriptor.m_Function));
+
+ activationDescriptor.m_Function = ActivationFunction::HardSwish;
+ BOOST_CHECK_MESSAGE((FuseActivationSimpleTest<Convolution2dTest<DataType::QAsymmU8>, DataType::QAsymmU8>
+ (activationDescriptor, Compute::GpuAcc)), "Convolution + Activation function " <<
+ static_cast<int>(activationDescriptor.m_Function));
+ BOOST_CHECK_MESSAGE((FuseActivationSimpleTest<FullyConnectedTest<DataType::QAsymmU8>, DataType::QAsymmU8>
+ (activationDescriptor, Compute::GpuAcc)), "FullyConnected + Activation function " <<
+ static_cast<int>(activationDescriptor.m_Function));
}
#endif
diff --git a/src/backends/neon/workloads/NeonFullyConnectedWorkload.cpp b/src/backends/neon/workloads/NeonFullyConnectedWorkload.cpp
index 39fb4c9..31489a0 100644
--- a/src/backends/neon/workloads/NeonFullyConnectedWorkload.cpp
+++ b/src/backends/neon/workloads/NeonFullyConnectedWorkload.cpp
@@ -27,16 +27,6 @@
const FullyConnectedDescriptor& descriptor,
const ActivationDescriptor* activationDescriptor)
{
- if (activationDescriptor)
- {
- std::vector<ActivationFunction> activations = {ActivationFunction::ReLu, ActivationFunction::BoundedReLu};
- if (std::find(activations.begin(), activations.end(), activationDescriptor->m_Function) == activations.end())
- {
- return arm_compute::Status{
- arm_compute::ErrorCode::RUNTIME_ERROR, "NeonFullyConnectedWorkload :Unsupported Activation Function"};
- }
- }
-
const arm_compute::TensorInfo aclInput = BuildArmComputeTensorInfo(input);
const arm_compute::TensorInfo aclOutput = BuildArmComputeTensorInfo(output);
const arm_compute::TensorInfo aclWeights = BuildArmComputeTensorInfo(weights);