Fused Depthwise Convolution Batch Normalization node. More...

#include <FusedDepthwiseConvolutionBatchNormalizationNode.h>

Collaboration diagram for FusedDepthwiseConvolutionBatchNormalizationNode:

Public Member Functions
	FusedDepthwiseConvolutionBatchNormalizationNode (float epsilon, PadStrideInfo info, unsigned int depth_multiplier, DepthwiseConvolutionMethod method, ActivationLayerInfo fused_activation=ActivationLayerInfo())
	Constructor. More...

void	set_depthwise_convolution_method (DepthwiseConvolutionMethod method)
	Sets the depthwise convolution layer method to use. More...

DepthwiseConvolutionMethod	depthwise_convolution_method () const
	Depthwise convolution layer method accessor. More...

float	epsilon () const
	Epsilon parameter accessor. More...

ActivationLayerInfo	fused_activation () const
	Returns fused activation. More...

void	set_fused_activation (ActivationLayerInfo fused_activation)
	Sets fused activation. More...

void	set_convolution_method (ConvolutionMethod method)
	Sets the convolution layer method to use. More...

unsigned int	depth_multiplier () const
	Depth multiplier accessor. More...

PadStrideInfo	convolution_info () const
	Convolution metadata accessor. More...

NodeType	type () const override
	Returns node's type. More...

bool	forward_descriptors () override
	Forwards descriptor information to outputs if possible. More...

TensorDescriptor	configure_output (size_t idx) const override
	Calculates output configuration. More...

void	accept (INodeVisitor &v) override
	Accepts a node visitor. More...

Public Member Functions inherited from INode
	INode ()
	Constructor. More...

virtual	~INode ()=default
	Destructor. More...

	INode (const INode &)=delete
	Prevent instances of this class from being copied (As this class contains pointers) More...

INode &	operator= (const INode &)=delete
	Prevent instances of this class from being copy assigned (As this class contains pointers) More...

	INode (INode &&)=default
	Allow instances of this class to be moved. More...

INode &	operator= (INode &&)=default
	Allow instances of this class to be move assigned. More...

virtual Status	validate () const
	Validate node. More...

std::string	name () const
	Returns node's name. More...

NodeID	id () const
	Returns node's ID. More...

const Graph *	graph () const
	Returns node's Graph. More...

Graph *	graph ()
	Returns node's Graph. More...

void	set_graph (Graph *g)
	Sets the graph that this node is registered to. More...

void	set_id (NodeID id)
	Sets the node id. More...

void	set_common_node_parameters (NodeParams common_params)
	Sets common node parameters. More...

void	set_requested_target (Target target)
	Sets target preference. More...

void	set_assigned_target (Target target)
	Sets the final execution target. More...

void	set_output_tensor (TensorID tid, size_t idx)
	Sets the output tensor of at a given index. More...

const std::vector< TensorID > &	inputs () const
	Returns inputs of the node. More...

const std::vector< TensorID > &	outputs () const
	Returns outputs of the node. More...

const std::vector< EdgeID > &	input_edges () const
	Returns input edge set. More...

const std::set< EdgeID > &	output_edges () const
	Returns output edge set. More...

TensorID	input_id (size_t idx) const
	Returns the tensor ID of a given input of the node. More...

TensorID	output_id (size_t idx) const
	Returns the tensor ID of a given output of the node. More...

Tensor *	input (size_t idx) const
	Returns the tensor of a given input of the node. More...

Tensor *	output (size_t idx) const
	Returns the tensor of a given output of the node. More...

EdgeID	input_edge_id (size_t idx) const
	Returns the edge ID of a given input of the node. More...

Edge *	input_edge (size_t idx) const
	Returns the edge of a given input of the node. More...

size_t	num_inputs () const
	Returns number of inputs of the node. More...

size_t	num_outputs () const
	Returns number of outputs of the node. More...

NodeParams	common_node_params () const
	Returns common node parameters. More...

Target	requested_target () const
	Returns requested target for this node. More...

Target	assigned_target () const
	Returns assigned target for this node. More...

Static Public Member Functions
static TensorDescriptor	compute_output_descriptor (const TensorDescriptor &input_descriptor, const TensorDescriptor &weights_descriptor, const PadStrideInfo &info, int depth_multiplier)
	Computes convolution output descriptor. More...

Static Public Attributes
static constexpr NodeType	node_type = NodeType::FusedDepthwiseConvolutionBatchNormalizationLayer

Detailed Description

Fused Depthwise Convolution Batch Normalization node.

Definition at line 34 of file FusedDepthwiseConvolutionBatchNormalizationNode.h.

Constructor & Destructor Documentation

◆ FusedDepthwiseConvolutionBatchNormalizationNode()

FusedDepthwiseConvolutionBatchNormalizationNode	(	float	epsilon,
		PadStrideInfo	info,
		unsigned int	depth_multiplier,
		DepthwiseConvolutionMethod	method,
		ActivationLayerInfo	fused_activation = `ActivationLayerInfo()`
	)

Constructor.

Parameters

[in]	epsilon	Epsilon parameter.
[in]	info	Convolution layer attributes.
[in]	depth_multiplier	(Optional) Multiplier to apply to the input's depth in order to retrieve the output's depth. Defaults to 1.
[in]	method	(Optional) Convolution method to use
[in]	fused_activation	(Optional) Fused activation layer. Disabled if not specified

Definition at line 35 of file FusedDepthwiseConvolutionBatchNormalizationNode.cpp.

     : _epsilon(epsilon),
       _info(std::move(info)),
       _depth_multiplier(depth_multiplier),
       _method(method),
       _fused_activation(fused_activation)
 {
     _input_edges.resize(7, EmptyEdgeID);
     _outputs.resize(1, NullTensorID);
 }

Member Function Documentation

◆ accept()

void accept ( INodeVisitor & v )

overridevirtual

Accepts a node visitor.

Parameters

[in] v Visitor to accept

Implements INode.

Definition at line 145 of file FusedDepthwiseConvolutionBatchNormalizationNode.cpp.

 {
     v.visit(*this);
 }

References INodeVisitor::visit().

◆ compute_output_descriptor()

TensorDescriptor compute_output_descriptor	(	const TensorDescriptor &	input_descriptor,
		const TensorDescriptor &	weights_descriptor,
		const PadStrideInfo &	info,
		int	depth_multiplier
	)

static

Computes convolution output descriptor.

Parameters

[in]	input_descriptor	Input descriptor
[in]	weights_descriptor	Weights descriptor
[in]	info	Convolution operation attributes
[in]	depth_multiplier	Depth multiplier

Returns: Output descriptor

Definition at line 88 of file FusedDepthwiseConvolutionBatchNormalizationNode.cpp.

 {
     unsigned int output_width  = 0;
     unsigned int output_height = 0;
  
     const unsigned int input_width    = get_dimension_size(input_descriptor, DataLayoutDimension::WIDTH);
     const unsigned int input_height   = get_dimension_size(input_descriptor, DataLayoutDimension::HEIGHT);
     const unsigned int input_channels = get_dimension_size(input_descriptor, DataLayoutDimension::CHANNEL);
     const unsigned int kernel_width   = get_dimension_size(weights_descriptor, DataLayoutDimension::WIDTH);
     const unsigned int kernel_height  = get_dimension_size(weights_descriptor, DataLayoutDimension::HEIGHT);
  
     std::tie(output_width, output_height) =
         scaled_dimensions(input_width, input_height, kernel_width, kernel_height, info);
  
     TensorDescriptor output_descriptor = input_descriptor;
     output_descriptor.shape.set(get_dimension_idx(output_descriptor.layout, DataLayoutDimension::WIDTH), output_width);
     output_descriptor.shape.set(get_dimension_idx(output_descriptor.layout, DataLayoutDimension::HEIGHT),
                                 output_height);
     output_descriptor.shape.set(get_dimension_idx(output_descriptor.layout, DataLayoutDimension::CHANNEL),
                                 input_channels * depth_multiplier);
  
     return output_descriptor;
 }

References arm_compute::CHANNEL, FusedDepthwiseConvolutionBatchNormalizationNode::depth_multiplier(), arm_compute::graph::get_dimension_idx(), arm_compute::graph::get_dimension_size(), arm_compute::HEIGHT, arm_compute::test::validation::info, TensorDescriptor::layout, arm_compute::scaled_dimensions(), TensorShape::set(), TensorDescriptor::shape, and arm_compute::WIDTH.

Referenced by FusedDepthwiseConvolutionBatchNormalizationNode::configure_output().

◆ configure_output()

TensorDescriptor configure_output ( size_t idx ) const

overridevirtual

Calculates output configuration.

Parameters

[in] idx Output index to configure

Returns: Output descriptor configuration

Implements INode.

Definition at line 127 of file FusedDepthwiseConvolutionBatchNormalizationNode.cpp.

 {
     ARM_COMPUTE_UNUSED(idx);
     const Tensor *src     = input(0);
     const Tensor *weights = input(1);
  
     ARM_COMPUTE_ERROR_ON(src == nullptr || weights == nullptr);
  
     TensorDescriptor output_info = compute_output_descriptor(src->desc(), weights->desc(), _info, _depth_multiplier);
  
     return output_info;
 }

References ARM_COMPUTE_ERROR_ON, ARM_COMPUTE_UNUSED, FusedDepthwiseConvolutionBatchNormalizationNode::compute_output_descriptor(), Tensor::desc(), INode::input(), arm_compute::test::validation::output_info, and arm_compute::test::validation::src.

Referenced by FusedDepthwiseConvolutionBatchNormalizationNode::forward_descriptors().

◆ convolution_info()

PadStrideInfo convolution_info ( ) const

Convolution metadata accessor.

Returns: Convolution information

Definition at line 67 of file FusedDepthwiseConvolutionBatchNormalizationNode.cpp.

 {
     return _info;
 }

Referenced by arm_compute::graph::backends::detail::create_fused_depthwise_convolution_batch_normalization_layer().

◆ depth_multiplier()

unsigned int depth_multiplier ( ) const

Depth multiplier accessor.

Returns: Depth multiplier

Definition at line 72 of file FusedDepthwiseConvolutionBatchNormalizationNode.cpp.

 {
     return _depth_multiplier;
 }

Referenced by FusedDepthwiseConvolutionBatchNormalizationNode::compute_output_descriptor(), and arm_compute::graph::backends::detail::create_fused_depthwise_convolution_batch_normalization_layer().

◆ depthwise_convolution_method()

DepthwiseConvolutionMethod depthwise_convolution_method ( ) const

Depthwise convolution layer method accessor.

Note: This is an indication on which depthwise convolution layer implementation to use, if it fails to be created the library's heuristic approach will be used

Returns: Depthwise convolution layer method to be used by the node

Definition at line 57 of file FusedDepthwiseConvolutionBatchNormalizationNode.cpp.

 {
     return _method;
 }

◆ epsilon()

float epsilon ( ) const

Epsilon parameter accessor.

Returns: Epsilon parameter

Definition at line 62 of file FusedDepthwiseConvolutionBatchNormalizationNode.cpp.

 {
     return _epsilon;
 }

Referenced by arm_compute::graph::backends::detail::create_fused_depthwise_convolution_batch_normalization_layer().

◆ forward_descriptors()

bool forward_descriptors ( )

overridevirtual

Forwards descriptor information to outputs if possible.

Returns: True if descriptor information could be forwarded otherwise false

Implements INode.

Definition at line 115 of file FusedDepthwiseConvolutionBatchNormalizationNode.cpp.

 {
     if ((input_id(0) != NullTensorID) && (input_id(1) != NullTensorID) && (output_id(0) != NullTensorID))
     {
         Tensor *dst = output(0);
         ARM_COMPUTE_ERROR_ON(dst == nullptr);
         dst->desc() = configure_output(0);
         return true;
     }
     return false;
 }

References ARM_COMPUTE_ERROR_ON, FusedDepthwiseConvolutionBatchNormalizationNode::configure_output(), arm_compute::test::validation::dst, INode::input_id(), arm_compute::graph::NullTensorID, INode::output(), and INode::output_id().

◆ fused_activation()

ActivationLayerInfo fused_activation ( ) const

Returns fused activation.

Returns: Fused activation

Definition at line 77 of file FusedDepthwiseConvolutionBatchNormalizationNode.cpp.

 {
     return _fused_activation;
 }

Referenced by arm_compute::graph::backends::detail::create_fused_depthwise_convolution_batch_normalization_layer(), and FusedDepthwiseConvolutionBatchNormalizationNode::set_fused_activation().

◆ set_convolution_method()

void set_convolution_method ( ConvolutionMethod method )

Sets the convolution layer method to use.

Parameters

[in] method Method to use for convolution

◆ set_depthwise_convolution_method()

void set_depthwise_convolution_method ( DepthwiseConvolutionMethod method )

Sets the depthwise convolution layer method to use.

Parameters

[in] method Method to use for depthwise convolution

Definition at line 51 of file FusedDepthwiseConvolutionBatchNormalizationNode.cpp.

 {
     _method = method;
 }

◆ set_fused_activation()

void set_fused_activation ( ActivationLayerInfo fused_activation )

Sets fused activation.

Parameters

[in] fused_activation Fused activation to set

Definition at line 82 of file FusedDepthwiseConvolutionBatchNormalizationNode.cpp.

 {
     _fused_activation = fused_activation;
 }

References FusedDepthwiseConvolutionBatchNormalizationNode::fused_activation().

◆ type()

NodeType type ( ) const

overridevirtual

Returns node's type.

Returns: Node's type

Implements INode.

Definition at line 140 of file FusedDepthwiseConvolutionBatchNormalizationNode.cpp.

 {
     return FusedDepthwiseConvolutionBatchNormalizationNode::node_type;
 }

References FusedDepthwiseConvolutionBatchNormalizationNode::node_type.

Referenced by arm_compute::graph::backends::detail::create_fused_depthwise_convolution_batch_normalization_layer().

Field Documentation

◆ node_type

constexpr NodeType node_type = NodeType::FusedDepthwiseConvolutionBatchNormalizationLayer

staticconstexpr

Definition at line 123 of file FusedDepthwiseConvolutionBatchNormalizationNode.h.

Referenced by FusedDepthwiseConvolutionBatchNormalizationNode::type().

The documentation for this class was generated from the following files:

arm_compute/graph/nodes/FusedDepthwiseConvolutionBatchNormalizationNode.h
src/graph/nodes/FusedDepthwiseConvolutionBatchNormalizationNode.cpp

Public Member Functions

Static Public Member Functions

Static Public Attributes

Detailed Description

Constructor & Destructor Documentation

◆ FusedDepthwiseConvolutionBatchNormalizationNode()

Member Function Documentation

◆ accept()

◆ compute_output_descriptor()

◆ configure_output()

◆ convolution_info()

◆ depth_multiplier()

◆ depthwise_convolution_method()

◆ epsilon()

◆ forward_descriptors()

◆ fused_activation()

◆ set_convolution_method()

◆ set_depthwise_convolution_method()

◆ set_fused_activation()

◆ type()

Field Documentation

◆ node_type