INetwork is the public interface to Network. More...

#include <nvneural/NetworkTypes.h>

Inheritance diagram for nvneural::INetwork:

Public Member Functions
virtual NeuralResult	attachNetworkDebugger (INetworkDebugger *pDebugger) noexcept=0
	Attaches a debugger object to the network. More...

virtual NeuralResult	attachReshapeDebugger (INetworkReshapeDebugger *pDebugger) noexcept=0
	Attaches a reshape debugger object to the network. More...

virtual NeuralResult	bindCurrentThread () noexcept=0
	Binds the network to the current thread. More...

virtual NeuralResult	clearLayers () noexcept=0
	Releases all layers from the network.

virtual NeuralResult	clearLoadedWeights () noexcept=0
	Clears the cache of loaded weights. More...

virtual NeuralResult	copyTensor (const ILayer pSourceLayer, ILayer pDestinationLayer) noexcept=0
	Copy the output tensor from one layer to another.

virtual IActivationFunction *	defaultForwardActivationFunction (const ILayer *pLayer) const noexcept=0
	Retrieves the activation function consumed by defaultForwardActivation. More...

virtual NeuralResult	detachAllNetworkDebuggers () noexcept=0
	Detaches all debuggers from the network.

virtual NeuralResult	detachAllReshapeDebuggers () noexcept=0
	Detaches all reshape debuggers from the network.

virtual NeuralResult	detachNetworkDebugger (const INetworkDebugger *pDebugger) noexcept=0
	Detaches a debugger object from the network.

virtual NeuralResult	detachReshapeDebugger (const INetworkReshapeDebugger *pDebugger) noexcept=0
	Detaches a debugger object from the network.

virtual const IClassRegistry *	getClassRegistry () const noexcept=0
	Get the IClassRegistry used by the network.

virtual NeuralResult	getNetworkDebuggers (INetworkDebuggerList **ppListOut) const noexcept=0
	Retrieves the list of attached network debuggers. More...

virtual NeuralResult	getProcessedLayerInputList (const ILayer pLayer, ILayerList *ppProcessedInputLayerList) const noexcept=0
	Get a layer input list and preprocess it to replace any inner NetworkLayer's layer by the NetworkLayer itself. More...

virtual NeuralResult	getWeightsNamesForLayer (IStringList *ppListOut, const ILayer pLayer) const noexcept=0
	Retrieves a list of weights consumed by a layer. More...

virtual NeuralResult	setClassRegistry (const IClassRegistry *pRegistry) noexcept=0
	Set an IClassRegistry object to this network. More...

virtual NeuralResult	synchronize () noexcept=0
	Performs a CPU/GPU sync and completes pending operations.

virtual NeuralResult	unload () noexcept=0
	Releases member data and theoretically breaks cycles between Networks/Layers not using weak references. More...

Initialization functions
virtual NeuralResult	attachBackend (INetworkBackend *pBackend) noexcept=0
	Attaches a backend to the network. More...

virtual NeuralResult	setDefaultBackendId (NetworkBackendId backendId) noexcept=0
	Sets the default backend for the network. More...

virtual NeuralResult	setDefaultTensorFormat (const TensorFormat &tensorFormat) noexcept=0
	Sets the default tensor format for the network.

virtual NeuralResult	setNetworkName (const char *pName) noexcept=0
	Sets the name for the network. More...

virtual NeuralResult	setOptimizationFlag (NetworkBackendId backendId, INetworkBackend::OptimizationCapability capability, bool enabled) noexcept=0
	Sets if the indicated optimization is applicable to the passed backend.

virtual bool	isOptimizationFlagEnabled (NetworkBackendId backendId, INetworkBackend::OptimizationCapability capability) const noexcept=0
	Gets if the indicated optimization is enabled to the passed backend.

virtual NeuralResult	setWeightsLoader (IWeightsLoader *pWeightsLoader) noexcept=0
	Sets the weights loader for the network. More...

virtual NeuralResult	pushLayer (ILayer *pLayer) noexcept=0
	Adds a layer to the network. More...

virtual NeuralResult	pushLayerAlias (const char pAliasName, ILayer pLayer) noexcept=0
	Adds an alias for a layer. More...

Reshape functions
virtual NeuralResult	reshapeAll () noexcept=0
	Reshapes the entire network.

virtual NeuralResult	reshape (ILayerList *pMarkedLayers) noexcept=0
	Reshapes a subgraph. More...

Inference
virtual NeuralResult	inferenceSubgraph (ILayerList pInputs, ILayerList pOutputs) noexcept=0
	Performs inference across a subgraph. More...

virtual NeuralResult	inference () noexcept=0
	Performs inference on the network. More...

Public Member Functions inherited from nvneural::INetworkRuntime
virtual NetworkBackendId	defaultBackendId () const noexcept=0
	Retrieves the default backend ID used by the network.

virtual NeuralResult	defaultForwardActivation (ILayer *pLayer) const noexcept=0
	Perform generic forward activation. More...

virtual TensorFormat	defaultTensorFormat () const noexcept=0
	Retrieves the default tensor format used by the network.

virtual const INetworkBackend *	getBackend (NetworkBackendId backendId) const noexcept=0
	This is an overloaded member function, provided for convenience. It differs from the above function only in what argument(s) it accepts.

virtual INetworkBackend *	getBackend (NetworkBackendId backendId) noexcept=0
	Retrieves an installed network backend. More...

virtual void *	getInternalTensor (const ILayer *pLayer, NetworkBackendId backendId) noexcept=0
	Retrieves internal tensor data for a layer. More...

virtual void *	getTensor (const ILayer pDataLayer, const ILayer pRequestingLayer, TensorFormat format) noexcept=0
	Retrieves tensor data from one layer on behalf of another. More...

virtual NeuralResult	getWeightsDimensionForLayer (TensorDimension pDimensionsOut, const ILayer pLayer, const char *pWeightsName, NetworkBackendId backendId, TensorFormat format) noexcept=0
	Retrieves the dimensions of specific weights data for a layer. More...

virtual const void *	getWeightsForLayer (const ILayer pLayer, const char pWeightsName, NetworkBackendId backendId, TensorFormat format) noexcept=0
	Retrieves specific weights data for a layer. More...

virtual NeuralResult	mapLayerAsWeightsData (const ILayer pDataLayer, const ILayer pWeightsLayer, const char *pWeightsName) noexcept=0
	Maps a layer as a source of weights data for another layer. More...

virtual const char *	networkName () const noexcept=0
	Retrieves the name of this network as a UTF-8 string.

virtual NeuralResult	reshapeLater (ILayer *pLayer) noexcept=0
	Queues a layer for later reshape. More...

virtual NeuralResult	setAffectedForward (ILayer *pLayer) noexcept=0
	Marks a layer and its descendants (those layers indirectly depending on it) as "affected.". More...

virtual IWeightsLoader *	weightsLoader () const noexcept=0
	Retrieves the IWeightsLoader implementation for this network.

virtual size_t	layerCount () const noexcept=0
	Gets the current number of layers in the network.

virtual ILayer *	getLayerByName (const char *pLayerName) const noexcept=0
	Retrieves a layer by name. More...

virtual NeuralResult	getAllLayers (ILayerList **ppListOut) const noexcept=0
	Retrieves a list of layers present in the network. More...

virtual NeuralResult	getInputLayers (ILayerList **ppListOut) const noexcept=0
	Retrieves a list of input layers present in the network. More...

virtual NeuralResult	getOutputLayers (ILayerList **ppListOut) const noexcept=0
	Retrieves a list of output layers present in the network. More...

virtual NeuralResult	getDependentLayers (ILayerList *ppListOut, const ILayer pLayer) const noexcept=0
	Retrieves a list of layers indirectly depending on pLayer. More...

virtual NeuralResult	getAcceptorLayers (ILayerList *ppListOut, const ILayer pLayer) const noexcept=0
	Retrieves a list of layers directly depending on pLayer. More...

virtual NeuralResult	getInfluenceLayers (ILayerList *ppListOut, const ILayer pLayer) const noexcept=0
	Retrieves a list of layers pLayer indirectly depends on. More...

Public Member Functions inherited from nvneural::IRefObject
virtual RefCount	addRef () const noexcept=0
	Increments the object's reference count. More...

virtual const void *	queryInterface (TypeId interface) const noexcept=0
	This is an overloaded member function, provided for convenience. It differs from the above function only in what argument(s) it accepts.

virtual void *	queryInterface (TypeId interface) noexcept=0
	Retrieves a new object interface pointer. More...

virtual RefCount	release () const noexcept=0
	Decrements the object's reference count and destroy the object if the reference count reaches zero. More...

Static Public Attributes
static const IRefObject::TypeId	typeID = 0x9ce8f71bf65288e9ul
	Interface TypeId for InterfaceOf purposes.

Static Public Attributes inherited from nvneural::INetworkRuntime
static const IRefObject::TypeId	typeID = 0x53191c073f990992ul
	Interface TypeId for InterfaceOf purposes.

Static Public Attributes inherited from nvneural::IRefObject
static const TypeId	typeID = 0x14ecc3f9de638e1dul
	Interface TypeId for InterfaceOf purposes.

Additional Inherited Members
Public Types inherited from nvneural::IRefObject
using	RefCount = std::uint32_t
	Typedef used to track the number of active references to an object.

using	TypeId = std::uint64_t
	Every interface must define a unique TypeId. This should be randomized.

Protected Member Functions inherited from nvneural::IRefObject
virtual	~IRefObject ()=default
	A protected destructor prevents accidental stack-allocation of IRefObjects or use with other smart pointer classes like std::unique_ptr.

Detailed Description

INetwork is the public interface to Network.

Member Function Documentation

◆ attachBackend()

virtual NeuralResult nvneural::INetwork::attachBackend ( INetworkBackend * pBackend )

pure virtualnoexcept

Attaches a backend to the network.

The backend is assumed to be fully constructed (all device handles and contexts initialized, etc.).

The backend ID must be unique within this network.

◆ attachNetworkDebugger()

virtual NeuralResult nvneural::INetwork::attachNetworkDebugger ( INetworkDebugger * pDebugger )

pure virtualnoexcept

Attaches a debugger object to the network.

The network takes a reference to the debugger.

◆ attachReshapeDebugger()

virtual NeuralResult nvneural::INetwork::attachReshapeDebugger ( INetworkReshapeDebugger * pDebugger )

pure virtualnoexcept

Attaches a reshape debugger object to the network.

The network takes a reference to the debugger.

◆ bindCurrentThread()

virtual NeuralResult nvneural::INetwork::bindCurrentThread ( )

pure virtualnoexcept

Binds the network to the current thread.

Some backends or external libraries may have thread affinity. Calling this function rebinds them for access from the calling thread.

Warning: For obvious reasons, this function is not thread-safe. If your code requires concurrent non-const access to the network from multiple threads, consider redesigning. Failing that, add your own synchronization around thread access; lock a mutex, rebind to current thread, do work, and unlock the mutex.

◆ clearLoadedWeights()

virtual NeuralResult nvneural::INetwork::clearLoadedWeights ( )

pure virtualnoexcept

Clears the cache of loaded weights.

Must be done if you change weights loader behavior significantly; there is no notification hook for the weights loader to trigger this itself.

◆ defaultForwardActivationFunction()

virtual IActivationFunction* nvneural::INetwork::defaultForwardActivationFunction ( const ILayer * pLayer ) const

pure virtualnoexcept

Retrieves the activation function consumed by defaultForwardActivation.

Parameters

pLayer Layer to query

Returns: IActivationFunction that would be used for activation, or nullptr if layer has no activation or if no valid object exists. Does not increment reference count.

◆ getNetworkDebuggers()

virtual NeuralResult nvneural::INetwork::getNetworkDebuggers ( INetworkDebuggerList ** ppListOut ) const

pure virtualnoexcept

Retrieves the list of attached network debuggers.

Parameters

ppListOut Pointer receiving a reference to a new INetworkDebuggerList object.

◆ getProcessedLayerInputList()

virtual NeuralResult nvneural::INetwork::getProcessedLayerInputList	(	const ILayer *	pLayer,
		ILayerList **	ppProcessedInputLayerList
	)		const

pure virtualnoexcept

Get a layer input list and preprocess it to replace any inner NetworkLayer's layer by the NetworkLayer itself.

Parameters

pLayer	the layer to retrive the input list from
ppProcessedInputLayerList	layer list containing the input of the layer

◆ getWeightsNamesForLayer()

virtual NeuralResult nvneural::INetwork::getWeightsNamesForLayer	(	IStringList **	ppListOut,
		const ILayer *	pLayer
	)		const

pure virtualnoexcept

Retrieves a list of weights consumed by a layer.

The list is sorted alphabetically. If the layer does not consume weights, an empty list is returned.

Parameters

ppListOut	Variable receiving a reference to a new IStringList. Caller must release the reference.
pLayer	Layer to query

◆ inference()

virtual NeuralResult nvneural::INetwork::inference ( )

pure virtualnoexcept

Performs inference on the network.

Layers not marked "affected," or not connected to the output layers returned by getOutputLayers, will not be evaluated.

◆ inferenceSubgraph()

virtual NeuralResult nvneural::INetwork::inferenceSubgraph	(	ILayerList *	pInputs,
		ILayerList *	pOutputs
	)

pure virtualnoexcept

Performs inference across a subgraph.

Parameters

pInputs	List of layers to perform inference from
pOutputs	List of layers to target inference to

◆ pushLayer()

virtual NeuralResult nvneural::INetwork::pushLayer ( ILayer * pLayer )

pure virtualnoexcept

Adds a layer to the network.

The layer is assumed to be fully constructed (all inputs hooked up, etc.).

◆ pushLayerAlias()

virtual NeuralResult nvneural::INetwork::pushLayerAlias	(	const char *	pAliasName,
		ILayer *	pLayer
	)

pure virtualnoexcept

Adds an alias for a layer.

The layer is assumed to be fully constructed (all inputs hooked up, etc.) and already added to the network.

◆ reshape()

virtual NeuralResult nvneural::INetwork::reshape ( ILayerList * pMarkedLayers )

pure virtualnoexcept

Reshapes a subgraph.

The specified layers and all layers indirectly depending on them are reshaped.

Parameters

pMarkedLayers List of layers in the subgraph

◆ setClassRegistry()

virtual NeuralResult nvneural::INetwork::setClassRegistry ( const IClassRegistry * pRegistry )

pure virtualnoexcept

Set an IClassRegistry object to this network.

Must be called before attaching network backends.

Parameters

pRegistry Pointer to an IClassRegistry object

◆ setDefaultBackendId()

virtual NeuralResult nvneural::INetwork::setDefaultBackendId ( NetworkBackendId backendId )

pure virtualnoexcept

Sets the default backend for the network.

The backend must already be known to the network (using attachBackend).

◆ setNetworkName()

virtual NeuralResult nvneural::INetwork::setNetworkName ( const char * pName )

pure virtualnoexcept

Sets the name for the network.

Network names may not contain the '/' character.

Parameters

pName Null-terminated UTF-8 string naming the network.

◆ setWeightsLoader()

virtual NeuralResult nvneural::INetwork::setWeightsLoader ( IWeightsLoader * pWeightsLoader )

pure virtualnoexcept

Sets the weights loader for the network.

Parameters

pWeightsLoader An object implementing IWeightsLoader

◆ unload()

virtual NeuralResult nvneural::INetwork::unload ( )

pure virtualnoexcept

Releases member data and theoretically breaks cycles between Networks/Layers not using weak references.

Allows for safe destruction of the Network instance.

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

Core/Inc/nvneural/NetworkTypes.h

Public Member Functions

Static Public Attributes

Additional Inherited Members

Detailed Description

Member Function Documentation

◆ attachBackend()

◆ attachNetworkDebugger()

◆ attachReshapeDebugger()

◆ bindCurrentThread()

◆ clearLoadedWeights()

◆ defaultForwardActivationFunction()

◆ getNetworkDebuggers()

◆ getProcessedLayerInputList()

◆ getWeightsNamesForLayer()

◆ inference()

◆ inferenceSubgraph()

◆ pushLayer()

◆ pushLayerAlias()

◆ reshape()

◆ setClassRegistry()

◆ setDefaultBackendId()

◆ setNetworkName()

◆ setWeightsLoader()

◆ unload()