patch/23.02/_convolution2d_layer_8cpp_source.xhtml

 //
 // Copyright © 2017,2022 Arm Ltd and Contributors. All rights reserved.
 // SPDX-License-Identifier: MIT
 //

 #include "Convolution2dLayer.hpp"
 #include "LayerCloneBase.hpp"

 #include <armnn/TypesUtils.hpp>

 #include <armnnUtils/DataLayoutIndexed.hpp>

 #include <armnn/backends/TensorHandle.hpp>
 #include <armnn/backends/WorkloadFactory.hpp>

 #include <string>

 using namespace armnnUtils;

 namespace armnn
 {

 Convolution2dLayer::Convolution2dLayer(const Convolution2dDescriptor& param, const char* name)
     : LayerWithParameters(param.GetNumInputs(), 1, LayerType::Convolution2d, param, name)
 {

 }

 void Convolution2dLayer::SerializeLayerParameters(ParameterStringifyFunction& fn) const
 {
     //using DescriptorType = Parameters;
     const std::vector<TensorShape>& inputShapes =
     {
         GetInputSlot(0).GetConnection()->GetTensorInfo().GetShape(),
         GetInputSlot(1).GetConnection()->GetTensorInfo().GetShape()
     };
     const TensorShape filterShape = inputShapes[1];
     DataLayoutIndexed dataLayoutIndex(m_Param.m_DataLayout);
     unsigned int filterWidth = filterShape[dataLayoutIndex.GetWidthIndex()];
     unsigned int filterHeight = filterShape[dataLayoutIndex.GetHeightIndex()];
     unsigned int outChannels = filterShape[0];

     fn("OutputChannels",std::to_string(outChannels));
     fn("FilterWidth",std::to_string(filterWidth));
     fn("FilterHeight",std::to_string(filterHeight));
     LayerWithParameters<Convolution2dDescriptor>::SerializeLayerParameters(fn);
 }

 std::unique_ptr<IWorkload> Convolution2dLayer::CreateWorkload(const IWorkloadFactory& factory) const
 {
     ARMNN_SCOPED_PROFILING_EVENT(Compute::Undefined, "Convolution2dLayer_CreateWorkload");
     Convolution2dQueueDescriptor descriptor;
     SetAdditionalInfo(descriptor);

     return factory.CreateWorkload(LayerType::Convolution2d, descriptor, PrepInfoAndDesc(descriptor));
 }

 Convolution2dLayer* Convolution2dLayer::Clone(Graph& graph) const
 {
     auto layer = CloneBase<Convolution2dLayer>(graph, m_Param, GetName());
     return std::move(layer);
 }

 std::vector<TensorShape> Convolution2dLayer::InferOutputShapes(const std::vector<TensorShape>& inputShapes) const
 {
     ARMNN_ASSERT(inputShapes.size() == 2);
     const TensorShape& inputShape = inputShapes[0];
     const TensorShape filterShape = inputShapes[1];

     // If we support multiple batch dimensions in the future, then this assert will need to change.
     ARMNN_ASSERT_MSG(inputShape.GetNumDimensions() == 4, "Convolutions will always have 4D input.");

     ARMNN_ASSERT( m_Param.m_StrideX > 0);
     ARMNN_ASSERT( m_Param.m_StrideY > 0);

     DataLayoutIndexed dataLayoutIndex(m_Param.m_DataLayout);

     unsigned int inWidth = inputShape[dataLayoutIndex.GetWidthIndex()];
     unsigned int inHeight = inputShape[dataLayoutIndex.GetHeightIndex()];
     unsigned int inBatchSize = inputShape[0];

     unsigned int filterWidth = filterShape[dataLayoutIndex.GetWidthIndex()];
     unsigned int dilatedFilterWidth = filterWidth + (m_Param.m_DilationX - 1) * (filterWidth - 1);
     unsigned int readWidth = (inWidth + m_Param.m_PadLeft + m_Param.m_PadRight) - dilatedFilterWidth;
     unsigned int outWidth = 1 + (readWidth / m_Param.m_StrideX);

     unsigned int filterHeight = filterShape[dataLayoutIndex.GetHeightIndex()];
     unsigned int dilatedFilterHeight = filterHeight + (m_Param.m_DilationY - 1) * (filterHeight - 1);
     unsigned int readHeight = (inHeight + m_Param.m_PadTop + m_Param.m_PadBottom) - dilatedFilterHeight;
     unsigned int outHeight = 1 + (readHeight / m_Param.m_StrideY);

     unsigned int outChannels = filterShape[0];
     unsigned int outBatchSize = inBatchSize;

     TensorShape tensorShape = m_Param.m_DataLayout == armnn::DataLayout::NHWC ?
         TensorShape( { outBatchSize, outHeight, outWidth, outChannels } ) :
         TensorShape( { outBatchSize, outChannels, outHeight, outWidth });

     return std::vector<TensorShape>({ tensorShape });
 }

 void Convolution2dLayer::ValidateTensorShapesFromInputs()
 {
     VerifyLayerConnections(m_Param.GetNumInputs(), CHECK_LOCATION());

     const TensorShape& outputShape = GetOutputSlot(0).GetTensorInfo().GetShape();

     VerifyShapeInferenceType(outputShape, m_ShapeInferenceMethod);

     ARMNN_ASSERT_MSG(GetInputSlot(1).GetConnection(),
                      "Convolution2dLayer: Weights should be connected to input slot 1.");

     std::vector<TensorShape> inferredShapes = InferOutputShapes({
              GetInputSlot(0).GetConnection()->GetTensorInfo().GetShape(),
              GetInputSlot(1).GetConnection()->GetTensorInfo().GetShape() });

     ARMNN_ASSERT(inferredShapes.size() == 1);

     ValidateAndCopyShape(outputShape, inferredShapes[0], m_ShapeInferenceMethod, "Convolution2dLayer");
 }

 Layer::ConstantTensors Convolution2dLayer::GetConstantTensorsByRef()
 {
     Layer::ConstantTensors tensors = GetConnectedConstantAsInputTensors();
     return tensors;
 }

 void Convolution2dLayer::ExecuteStrategy(IStrategy& strategy) const
 {
     strategy.ExecuteStrategy(this, GetParameters(), {}, GetName());
 }

 } // namespace armnn
armnn::Convolution2dDescriptor::m_PadBottom
uint32_t m_PadBottom
Padding bottom value in the height dimension.
Definition: Descriptors.hpp:540

-armnn::Convolution2dDescriptor::m_DataLayout
DataLayout m_DataLayout
The data layout to be used (NCHW, NHWC).
Definition: Descriptors.hpp:552

-DataLayoutIndexed.hpp

-armnn::LayerWithParameters< Convolution2dDescriptor >::m_Param
Convolution2dDescriptor m_Param
The parameters for the layer (not including tensor-valued weights etc.).
Definition: LayerWithParameters.hpp:52

-armnnUtils::DataLayoutIndexed::GetWidthIndex
unsigned int GetWidthIndex() const
Definition: DataLayoutIndexed.hpp:25

-armnn::Convolution2dDescriptor::GetNumInputs
uint32_t GetNumInputs() const
Definition: Descriptors.cpp:443

-armnn::TensorInfo::GetShape
const TensorShape & GetShape() const
Definition: Tensor.hpp:191

-TensorHandle.hpp

-WorkloadFactory.hpp

-armnn::IWorkloadFactory
Definition: WorkloadFactory.hpp:22

-armnn::Convolution2dDescriptor
A Convolution2dDescriptor for the Convolution2dLayer.
Definition: Descriptors.hpp:502

-armnn::LayerWithParameters::SerializeLayerParameters
void SerializeLayerParameters(ParameterStringifyFunction &fn) const override
Helper to serialize the layer parameters to string (currently used in DotSerializer and company)...
Definition: LayerWithParameters.hpp:23

-armnn::Convolution2dDescriptor::m_PadRight
uint32_t m_PadRight
Padding right value in the width dimension.
Definition: Descriptors.hpp:536

-armnn::Convolution2dLayer::Convolution2dLayer
Convolution2dLayer(const Convolution2dDescriptor &param, const char *name)
Constructor to create a Convolution2dLayer.
Definition: Convolution2dLayer.cpp:23

-armnn::Layer::VerifyShapeInferenceType
void VerifyShapeInferenceType(const TensorShape &outputShape, ShapeInferenceMethod shapeInferenceMethod)
Definition: Layer.cpp:491

-armnn
Copyright (c) 2021 ARM Limited and Contributors.
Definition: 01_00_quick_start.dox:6

-armnn::LayerWithParameters< Convolution2dDescriptor >::GetParameters
const Convolution2dDescriptor & GetParameters() const override
Definition: LayerWithParameters.hpp:19

-armnn::TensorShape
Definition: Tensor.hpp:20

-armnn::InputSlot::GetConnection
const IOutputSlot * GetConnection() const override
Definition: Layer.hpp:206

-armnn::Convolution2dDescriptor::m_DilationY
uint32_t m_DilationY
Dilation along y axis.
Definition: Descriptors.hpp:548

-armnn::Layer::ValidateAndCopyShape
void ValidateAndCopyShape(const TensorShape &outputShape, const TensorShape &inferredShape, const ShapeInferenceMethod shapeInferenceMethod, const std::string &layerName, const unsigned int outputSlotIndex=0)
Definition: Layer.cpp:422

-ARMNN_SCOPED_PROFILING_EVENT
#define ARMNN_SCOPED_PROFILING_EVENT(backendId, name)
Definition: Profiling.hpp:220

-armnnUtils::DataLayoutIndexed::GetHeightIndex
unsigned int GetHeightIndex() const
Definition: DataLayoutIndexed.hpp:24

-armnn::IStrategy
Definition: IStrategy.hpp:16

-armnn::Layer::VerifyLayerConnections
void VerifyLayerConnections(unsigned int expectedConnections, const CheckLocation &location) const
Definition: Layer.cpp:378

-armnn::Convolution2dDescriptor::m_PadTop
uint32_t m_PadTop
Padding top value in the height dimension.
Definition: Descriptors.hpp:538

-armnn::Layer::GetInputSlot
const InputSlot & GetInputSlot(unsigned int index) const override
Get a const input slot handle by slot index.
Definition: Layer.hpp:324

-armnn::Convolution2dDescriptor::m_StrideX
uint32_t m_StrideX
Stride value when proceeding through input for the width dimension.
Definition: Descriptors.hpp:542

-armnn::Convolution2dLayer::ValidateTensorShapesFromInputs
void ValidateTensorShapesFromInputs() override
Check if the input tensor shape(s) will lead to a valid configuration of Convolution2dLayer.
Definition: Convolution2dLayer.cpp:102

-TypesUtils.hpp

-armnn::Convolution2dLayer::ExecuteStrategy
void ExecuteStrategy(IStrategy &strategy) const override
Apply a visitor to this layer.
Definition: Convolution2dLayer.cpp:128

-armnn::GetNumInputs
uint32_t GetNumInputs(bool biasEnabled)
Definition: Descriptors.cpp:428

-armnn::IConnectableLayer::ConstantTensors
std::vector< std::reference_wrapper< std::shared_ptr< ConstTensorHandle > >> ConstantTensors
Definition: INetwork.hpp:124

-ARMNN_ASSERT_MSG
#define ARMNN_ASSERT_MSG(COND, MSG)
Definition: Assert.hpp:15

-armnnUtils::DataLayoutIndexed
Provides access to the appropriate indexes for Channels, Height and Width based on DataLayout...
Definition: DataLayoutIndexed.hpp:17

-armnn::Compute::Undefined

-Convolution2dLayer.hpp

-armnn::LayerType::Convolution2d

-ARMNN_ASSERT
#define ARMNN_ASSERT(COND)
Definition: Assert.hpp:14

-armnn::Convolution2dLayer::Clone
Convolution2dLayer * Clone(Graph &graph) const override
Creates a dynamically-allocated copy of this layer.
Definition: Convolution2dLayer.cpp:58

-armnn::IStrategy::ExecuteStrategy
virtual void ExecuteStrategy(const IConnectableLayer *layer, const armnn::BaseDescriptor &descriptor, const std::vector< armnn::ConstTensor > &constants, const char *name, const armnn::LayerBindingId id=0)=0

-armnn::Convolution2dLayer::GetConstantTensorsByRef
ConstantTensors GetConstantTensorsByRef() override
Retrieve the handles to the constant values connected to the layer.
Definition: Convolution2dLayer.cpp:122

-CHECK_LOCATION
#define CHECK_LOCATION()
Definition: Exceptions.hpp:203

-armnn::Convolution2dDescriptor::m_StrideY
uint32_t m_StrideY
Stride value when proceeding through input for the height dimension.
Definition: Descriptors.hpp:544

-armnn::Graph
Definition: Graph.hpp:30

-armnn::Layer::SetAdditionalInfo
void SetAdditionalInfo(QueueDescriptor &descriptor) const
Definition: Layer.cpp:274

-armnn::Convolution2dQueueDescriptor
Definition: WorkloadData.hpp:205

-armnn::Convolution2dDescriptor::m_DilationX
uint32_t m_DilationX
Dilation along x axis.
Definition: Descriptors.hpp:546

-armnn::Convolution2dLayer::CreateWorkload
virtual std::unique_ptr< IWorkload > CreateWorkload(const IWorkloadFactory &factory) const override
Makes a workload for the Convolution2d type.
Definition: Convolution2dLayer.cpp:49

-armnn::Convolution2dLayer::SerializeLayerParameters
void SerializeLayerParameters(ParameterStringifyFunction &fn) const override
Helper to serialize the layer parameters to string.
Definition: Convolution2dLayer.cpp:29

-armnn::LayerWithParameters< Convolution2dDescriptor >::GetConnectedConstantAsInputTensors
Layer::ConstantTensors GetConnectedConstantAsInputTensors()
Definition: LayerWithParameters.hpp:59

-armnn::LayerWithParameters< Convolution2dDescriptor >::PrepInfoAndDesc
WorkloadInfo PrepInfoAndDesc(QueueDescriptor &descriptor) const
Helper function to reduce duplication in *Layer::CreateWorkload.
Definition: LayerWithParameters.hpp:44

-armnnUtils
Definition: CompatibleTypes.hpp:10

-armnn::Layer::GetOutputSlot
const OutputSlot & GetOutputSlot(unsigned int index=0) const override
Get the const output slot handle by slot index.
Definition: Layer.hpp:326

-armnn::IOutputSlot::GetTensorInfo
virtual const TensorInfo & GetTensorInfo() const =0

-armnn::Layer::GetName
const char * GetName() const override
Returns the name of the layer.
Definition: Layer.hpp:319

-armnn::Convolution2dLayer
This layer represents a convolution 2d operation.
Definition: Convolution2dLayer.hpp:15

-armnn::LayerWithParameters
Definition: LayerWithParameters.hpp:14

-armnn::Convolution2dLayer::InferOutputShapes
std::vector< TensorShape > InferOutputShapes(const std::vector< TensorShape > &inputShapes) const override
By default returns inputShapes if the number of inputs are equal to number of outputs, otherwise infers the output shapes from given input shapes and layer properties.
Definition: Convolution2dLayer.cpp:64

-armnn::ParameterStringifyFunction
std::function< void(const std::string &name, const std::string &value)> ParameterStringifyFunction
Definition: SerializeLayerParameters.hpp:14

-armnn::IWorkloadFactory::CreateWorkload
virtual std::unique_ptr< IWorkload > CreateWorkload(LayerType type, const QueueDescriptor &descriptor, const WorkloadInfo &info) const
Definition: WorkloadFactory.cpp:1561

-armnn::OutputSlot::GetTensorInfo
const TensorInfo & GetTensorInfo() const override
Definition: Layer.cpp:92

-armnn::Layer::m_ShapeInferenceMethod
ShapeInferenceMethod m_ShapeInferenceMethod
Definition: Layer.hpp:423

-armnn::Convolution2dDescriptor::m_PadLeft
uint32_t m_PadLeft
Padding left value in the width dimension.
Definition: Descriptors.hpp:534

-armnn::DataLayout::NHWC

-LayerCloneBase.hpp

-armnn::LayerType
LayerType
When adding a new layer, adapt also the LastLayer enum value in the enum class LayerType below...
Definition: Types.hpp:466

+Go to the documentation of this file. //

+ // Copyright © 2017,2022 Arm Ltd and Contributors. All rights reserved.

+ // SPDX-License-Identifier: MIT

+ //

+

+ #include "Convolution2dLayer.hpp"

+ #include "LayerCloneBase.hpp"

+

+ #include <armnn/TypesUtils.hpp>

+

+ #include <armnnUtils/DataLayoutIndexed.hpp>

+

+ #include <armnn/backends/TensorHandle.hpp>

+ #include <armnn/backends/WorkloadFactory.hpp>

+

+ #include <string>

+

+ using namespace armnnUtils;

+

+ namespace armnn

+ {

+

+ Convolution2dLayer::Convolution2dLayer(const Convolution2dDescriptor& param, const char* name)

+     : LayerWithParameters(param.GetNumInputs(), 1, LayerType::Convolution2d, param, name)

+ {

+

+ }

+

+ void Convolution2dLayer::SerializeLayerParameters(ParameterStringifyFunction& fn) const

+ {

+     //using DescriptorType = Parameters;

+     const std::vector<TensorShape>& inputShapes =

+     {

+         GetInputSlot(0).GetConnection()->GetTensorInfo().GetShape(),

+         GetInputSlot(1).GetConnection()->GetTensorInfo().GetShape()

+     };

+     const TensorShape filterShape = inputShapes[1];

+     DataLayoutIndexed dataLayoutIndex(m_Param.m_DataLayout);

+     unsigned int filterWidth = filterShape[dataLayoutIndex.GetWidthIndex()];

+     unsigned int filterHeight = filterShape[dataLayoutIndex.GetHeightIndex()];

+     unsigned int outChannels = filterShape[0];

+

+     fn("OutputChannels",std::to_string(outChannels));

+     fn("FilterWidth",std::to_string(filterWidth));

+     fn("FilterHeight",std::to_string(filterHeight));

+     LayerWithParameters<Convolution2dDescriptor>::SerializeLayerParameters(fn);

+ }

+

+ std::unique_ptr<IWorkload> Convolution2dLayer::CreateWorkload(const IWorkloadFactory& factory) const

+ {

+     ARMNN_SCOPED_PROFILING_EVENT(Compute::Undefined, "Convolution2dLayer_CreateWorkload");

+     Convolution2dQueueDescriptor descriptor;

+     SetAdditionalInfo(descriptor);

+

+     return factory.CreateWorkload(LayerType::Convolution2d, descriptor, PrepInfoAndDesc(descriptor));

+ }

+

+ Convolution2dLayer* Convolution2dLayer::Clone(Graph& graph) const

+ {

+     auto layer = CloneBase<Convolution2dLayer>(graph, m_Param, GetName());

+     return std::move(layer);

+ }

+

+ std::vector<TensorShape> Convolution2dLayer::InferOutputShapes(const std::vector<TensorShape>& inputShapes) const

+ {

+     ARMNN_ASSERT(inputShapes.size() == 2);

+     const TensorShape& inputShape = inputShapes[0];

+     const TensorShape filterShape = inputShapes[1];

+

+     // If we support multiple batch dimensions in the future, then this assert will need to change.

+     ARMNN_ASSERT_MSG(inputShape.GetNumDimensions() == 4, "Convolutions will always have 4D input.");

+

+     ARMNN_ASSERT( m_Param.m_StrideX > 0);

+     ARMNN_ASSERT( m_Param.m_StrideY > 0);

+

+     DataLayoutIndexed dataLayoutIndex(m_Param.m_DataLayout);

+

+     unsigned int inWidth = inputShape[dataLayoutIndex.GetWidthIndex()];

+     unsigned int inHeight = inputShape[dataLayoutIndex.GetHeightIndex()];

+     unsigned int inBatchSize = inputShape[0];

+

+     unsigned int filterWidth = filterShape[dataLayoutIndex.GetWidthIndex()];

+     unsigned int dilatedFilterWidth = filterWidth + (m_Param.m_DilationX - 1) * (filterWidth - 1);

+     unsigned int readWidth = (inWidth + m_Param.m_PadLeft + m_Param.m_PadRight) - dilatedFilterWidth;

+     unsigned int outWidth = 1 + (readWidth / m_Param.m_StrideX);

+

+     unsigned int filterHeight = filterShape[dataLayoutIndex.GetHeightIndex()];

+     unsigned int dilatedFilterHeight = filterHeight + (m_Param.m_DilationY - 1) * (filterHeight - 1);

+     unsigned int readHeight = (inHeight + m_Param.m_PadTop + m_Param.m_PadBottom) - dilatedFilterHeight;

+     unsigned int outHeight = 1 + (readHeight / m_Param.m_StrideY);

+

+     unsigned int outChannels = filterShape[0];

+     unsigned int outBatchSize = inBatchSize;

+

+     TensorShape tensorShape = m_Param.m_DataLayout == armnn::DataLayout::NHWC ?

+         TensorShape( { outBatchSize, outHeight, outWidth, outChannels } ) :

+         TensorShape( { outBatchSize, outChannels, outHeight, outWidth });

+

+     return std::vector<TensorShape>({ tensorShape });

+ }

+

+ void Convolution2dLayer::ValidateTensorShapesFromInputs()

+ {

+     VerifyLayerConnections(m_Param.GetNumInputs(), CHECK_LOCATION());

+

+     const TensorShape& outputShape = GetOutputSlot(0).GetTensorInfo().GetShape();

+

+     VerifyShapeInferenceType(outputShape, m_ShapeInferenceMethod);

+

+     ARMNN_ASSERT_MSG(GetInputSlot(1).GetConnection(),

+                      "Convolution2dLayer: Weights should be connected to input slot 1.");

+

+     std::vector<TensorShape> inferredShapes = InferOutputShapes({

+              GetInputSlot(0).GetConnection()->GetTensorInfo().GetShape(),

+              GetInputSlot(1).GetConnection()->GetTensorInfo().GetShape() });

+

+     ARMNN_ASSERT(inferredShapes.size() == 1);

+

+     ValidateAndCopyShape(outputShape, inferredShapes[0], m_ShapeInferenceMethod, "Convolution2dLayer");

+ }

+

+ Layer::ConstantTensors Convolution2dLayer::GetConstantTensorsByRef()

+ {

+     Layer::ConstantTensors tensors = GetConnectedConstantAsInputTensors();

+     return tensors;

+ }

+

+ void Convolution2dLayer::ExecuteStrategy(IStrategy& strategy) const

+ {

+     strategy.ExecuteStrategy(this, GetParameters(), {}, GetName());

+ }

+

+ } // namespace armnn