a network definition for input to the builder More...

#include <NvInfer.h>

Public Member Functions
virtual ITensor *	addInput (const char *name, DataType type, Dims dimensions)=0
	add an input tensor to the network More...

virtual void	markOutput (ITensor &tensor)=0
	mark a tensor as a network output More...

virtual IConvolutionLayer *	addConvolution (ITensor &input, int nbOutputMaps, DimsHW kernelSize, Weights kernelWeights, Weights biasWeights)=0
	add a convolution layer to the network More...

virtual IFullyConnectedLayer *	addFullyConnected (ITensor &input, int nbOutputs, Weights kernelWeights, Weights biasWeights)=0
	add a fully connected layer to the network More...

virtual IActivationLayer *	addActivation (ITensor &input, ActivationType type)=0
	add an activation layer to the network More...

virtual IPoolingLayer *	addPooling (ITensor &input, PoolingType type, DimsHW windowSize)=0
	add a pooling layer to the network More...

virtual ILRNLayer *	addLRN (ITensor &input, int window, float alpha, float beta, float k)=0
	add a LRN layer to the network More...

virtual IScaleLayer *	addScale (ITensor &input, ScaleMode mode, Weights shift, Weights scale, Weights power)=0
	add a Scale layer to the network More...

virtual ISoftMaxLayer *	addSoftMax (ITensor &input)=0
	add a Scale layer to the network More...

virtual IConcatenationLayer *	addConcatenation (ITensor const inputs, int nbInputs)=0
	add a concatenation layer to the network More...

virtual IDeconvolutionLayer *	addDeconvolution (ITensor &input, int nbOutputMaps, DimsHW kernelSize, Weights kernelWeights, Weights biasWeights)=0
	add a deconvolution layer to the network More...

virtual IElementWiseLayer *	addElementWise (ITensor &input1, ITensor &input2, ElementWiseOperation op)=0
	add an elementwise layer to the network More...

virtual IRNNLayer *	addRNN (ITensor &inputs, int layerCount, std::size_t hiddenSize, int maxSeqLen, RNNOperation op, RNNInputMode mode, RNNDirection dir, Weights weights, Weights bias)=0
	add an `layerCount` deep RNN layer to the network with a sequence length of `maxSeqLen` and `hiddenSize` internal state per layer. More...

virtual IPluginLayer *	addPlugin (ITensor const inputs, int nbInputs, IPlugin &plugin)=0
	add a plugin layer to the network More...

virtual IUnaryLayer *	addUnary (ITensor &input, UnaryOperation operation)=0
	Add a unary layer to the network. More...

virtual IPaddingLayer *	addPadding (ITensor &input, DimsHW prePadding, DimsHW postPadding)=0
	Add a padding layer to the network. More...

virtual IShuffleLayer *	addShuffle (ITensor &input)=0
	add a shuffle layer to the network More...

virtual void	setPoolingOutputDimensionsFormula (IOutputDimensionsFormula *formula)=0
	set the pooling output dimensions formula More...

virtual IOutputDimensionsFormula &	getPoolingOutputDimensionsFormula () const =0
	get the pooling output dimensions formula More...

virtual void	setConvolutionOutputDimensionsFormula (IOutputDimensionsFormula *formula)=0
	set the convolution output dimensions formula More...

virtual IOutputDimensionsFormula &	getConvolutionOutputDimensionsFormula () const =0
	get the convolution output dimensions formula More...

virtual void	setDeconvolutionOutputDimensionsFormula (IOutputDimensionsFormula *formula)=0
	set the deconvolution output dimensions formula More...

virtual IOutputDimensionsFormula &	getDeconvolutionOutputDimensionsFormula () const =0
	get the deconvolution output dimensions formula More...

virtual int	getNbLayers () const =0
	get the number of layers in the network More...

virtual ILayer *	getLayer (int index) const =0
	get the layer specified by the given index More...

virtual int	getNbInputs () const =0
	get the number of inputs in the network More...

virtual ITensor *	getInput (int index) const =0
	get the input tensor specified by the given index More...

virtual int	getNbOutputs () const =0
	get the number of outputs in the network More...

virtual ITensor *	getOutput (int index) const =0
	get the output tensor specified by the given index More...

virtual void	destroy ()=0
	destroy this INetworkDefinition object

Detailed Description

a network definition for input to the builder

Member Function Documentation

virtual IActivationLayer* nvinfer1::INetworkDefinition::addActivation	(	ITensor &	input,
		ActivationType	type
	)

pure virtual

add an activation layer to the network

Parameters

input	the input tensor to the layer
type	the type of activation function to apply

See Also: IActivationLayer ActivationType

Returns: the new convolution layer, or null if it could not be created

virtual IConcatenationLayer* nvinfer1::INetworkDefinition::addConcatenation	(	ITensor const	inputs,
		int	nbInputs
	)

pure virtual

add a concatenation layer to the network

Parameters

inputs	the input tensors to the layer
nbInputs	the number of input tensors

See Also: IConcatenationLayer

Returns: the new concatenation layer, or null if it could not be created

Warning: All tensors must have the same dimensions for all dimensions except for channel.

virtual IConvolutionLayer* nvinfer1::INetworkDefinition::addConvolution	(	ITensor &	input,
		int	nbOutputMaps,
		DimsHW	kernelSize,
		Weights	kernelWeights,
		Weights	biasWeights
	)

pure virtual

add a convolution layer to the network

Parameters

input	the input tensor to the convolution
nbOutputMaps	the number of output feature maps for the convolution
kernelSize	the HW-dimensions of the convolution kernel
kernelWeights	the kernel weights for the convolution
biasWeights	the optional bias weights for the convolution

See Also: IConvolutionLayer

Returns: the new convolution layer, or null if it could not be created

virtual IDeconvolutionLayer* nvinfer1::INetworkDefinition::addDeconvolution	(	ITensor &	input,
		int	nbOutputMaps,
		DimsHW	kernelSize,
		Weights	kernelWeights,
		Weights	biasWeights
	)

pure virtual

add a deconvolution layer to the network

Parameters

input	the input tensor to the layer
nbOutputMaps	the number of output feature maps
kernelSize	the HW-dimensions of the convolution kernel
kernelWeights	the kernel weights for the convolution
biasWeights	the optional bias weights for the convolution

See Also: IDeconvolutionLayer

Returns: the new deconvolution layer, or null if it could not be created

virtual IElementWiseLayer* nvinfer1::INetworkDefinition::addElementWise	(	ITensor &	input1,
		ITensor &	input2,
		ElementWiseOperation	op
	)

pure virtual

add an elementwise layer to the network

Parameters

input1	the first input tensor to the layer
input2	the second input tensor to the layer
op	the binary operation that the layer applies

See Also: IElementWiseLayer

Returns: the new elementwise layer, or null if it could not be created

virtual IFullyConnectedLayer* nvinfer1::INetworkDefinition::addFullyConnected	(	ITensor &	input,
		int	nbOutputs,
		Weights	kernelWeights,
		Weights	biasWeights
	)

pure virtual

add a fully connected layer to the network

Parameters

input	the input tensor to the layer
nbOutputs	the number of outputs of the layer
kernelWeights	the kernel weights for the convolution
biasWeights	the optional bias weights for the convolution

See Also: IFullyConnectedLayer

the input to a fully connected layer is automatically flattened to a tensor of the form NxCx1x1, and the output is of the form NxKx1x1, where C is the nunber of input activations per image, and K is the number of outputs per image.

Returns: the new convolution layer, or null if it could not be created

virtual ITensor* nvinfer1::INetworkDefinition::addInput	(	const char *	name,
		DataType	type,
		Dims	dimensions
	)

pure virtual

add an input tensor to the network

the name of the input tensor is used to find the index into the buffer array for an engine built from the network

Parameters

name	the name of the tensor
type	the type of the data held in the tensor
dimensions	the dimensions of the tensor

Only DataType::kFLOAT and DataType::kHALF are valid input tensor types. The volume of the dimension, including the maximum batch size, must be less than 2^30 elements.

See Also: ITensor

Returns: the new tensor or nullptr if there is an error

virtual ILRNLayer* nvinfer1::INetworkDefinition::addLRN	(	ITensor &	input,
		int	window,
		float	alpha,
		float	beta,
		float	k
	)

pure virtual

add a LRN layer to the network

Parameters

input	the input tensor to the layer
window	the size of the window
alpha	the alpha value for the LRN computation
beta	the beta value for the LRN computation
k	the k value for the LRN computation

See Also: ILRNLayer

Returns: the new LRN layer, or null if it could not be created

virtual IPaddingLayer* nvinfer1::INetworkDefinition::addPadding	(	ITensor &	input,
		DimsHW	prePadding,
		DimsHW	postPadding
	)

pure virtual

Add a padding layer to the network.

Parameters

input	the input tensor to the layer
prePadding	the padding to apply to the start of the tensor
postPadding	the padding to apply to the end of the tensor

See Also: IPaddingLayer

Returns: the new padding layer, or null if it could not be created

virtual IPluginLayer* nvinfer1::INetworkDefinition::addPlugin	(	ITensor const	inputs,
		int	nbInputs,
		IPlugin &	plugin
	)

pure virtual

add a plugin layer to the network

Parameters

inputs	the input tensors to the layer
nbInputs	the number of input tensors
plugin	the layer plugin

See Also: IPluginLayer

Returns: the new plugin layer, or null if it could not be created

virtual IPoolingLayer* nvinfer1::INetworkDefinition::addPooling	(	ITensor &	input,
		PoolingType	type,
		DimsHW	windowSize
	)

pure virtual

add a pooling layer to the network

Parameters

input	the input tensor to the layer
type	the type of pooling to apply
windowSize	the size of the pooling window

See Also: IPoolingLayer PoolingType

Returns: the new pooling layer, or null if it could not be created

virtual IRNNLayer* nvinfer1::INetworkDefinition::addRNN	(	ITensor &	inputs,
		int	layerCount,
		std::size_t	hiddenSize,
		int	maxSeqLen,
		RNNOperation	op,
		RNNInputMode	mode,
		RNNDirection	dir,
		Weights	weights,
		Weights	bias
	)

pure virtual

add an layerCount deep RNN layer to the network with a sequence length of maxSeqLen and hiddenSize internal state per layer.

Parameters

inputs	the input tensor to the layer.
layerCount	the number of layers in the RNN.
hiddenSize	the size of the internal hidden state for each layer.
maxSeqLen	the maximum length of the time sequence.
op	the type of RNN to execute.
mode	the input mode for the RNN.
dir	the direction to run the RNN.
weights	the weights for the weight matrix parameters of the RNN.
bias	the weights for the bias vectors parameters of the RNN.

The layout for the input tensor is {1, T, N, C} and defined as follows:

T - The number of time sequences to be executed.
N - The number of mini-batches for each time sequence.
C - The size of data to be submitted to the RNN.

The input tensors must be of the type DataType::kFLOAT or DataType::kHALF. The layout of the weights is row major and must be the same datatype as the input tensor. weights contain 8 matrices and bias contains 8 vectors.

The RNN layer outputs up to three tensors. The first tensor has dim {1, T, N, C}, is the output of the RNN for each timestep, and defined as follows:

T - The number of sequences to be executed.
N - The number of mini-batches for each time sequence.
C - The hidden state for each layer. Equal to getHiddenSize() if getDirection is RNNDirection::kUNIDIRECTION, and 2x getHiddenSize() otherwise.

The second tensor has dimension {1, L, N, H}, is the final hidden state of the RNN, and defined as follows.

L - The number of layers in the RNN, equal to getLayerCount()
N - The number of mini-batches for each time sequence.
H - The hidden state for each layer. Equal to getHiddenSize() if getDirection is RNNDirection::kUNIDIRECTION, and 2x getHiddenSize() otherwise.

The third tensor has dimension {1, L, N, H}, is the final cell state of the RNN, and defined as follows.

L - The number of layers in the RNN, equal to getLayerCount()
N - The number of mini-batches for each time sequence.
H - The hidden state for each layer. Equal to getHiddenSize() if getDirection is RNNDirection::kUNIDIRECTION, and 2x getHiddenSize() otherwise.

The third tensor is only available if getOperation() is RNNDirection::kLSTM.

Returns: the new RNN layer, or null if it could not be created.

See Also: IRNNLayer::setW(), IRNNLayer::setU(), IRNNLayer::setB(); IRNNLayer

virtual IScaleLayer* nvinfer1::INetworkDefinition::addScale	(	ITensor &	input,
		ScaleMode	mode,
		Weights	shift,
		Weights	scale,
		Weights	power
	)

pure virtual

add a Scale layer to the network

Parameters

input	the input tensor to the layer
mode	the scaling mode
shift	the shift value
scale	the scale value
power	the power value

If the weights are available, then the size of weights are dependent on the on the ScaleMode. For ScaleMode::kUNIFORM, the number of weights is equal to 1. For ScaleMode::kCHANNEL, the number of weights is equal to the channel dimension. For ScaleMode::kELEMENTWISE, the number of weights is equal to the volume of the input.

See Also: IScaleLayer

Returns: the new Scale layer, or null if it could not be created

virtual IShuffleLayer* nvinfer1::INetworkDefinition::addShuffle ( ITensor & input )

pure virtual

add a shuffle layer to the network

Parameters

input the input tensor to the layer

Returns: the new shuffle layer, or null if it could not be created

virtual ISoftMaxLayer* nvinfer1::INetworkDefinition::addSoftMax ( ITensor & input )

pure virtual

add a Scale layer to the network

See Also: ISoftMaxLayer

Returns: the new SoftMax layer, or null if it could not be created

virtual IUnaryLayer* nvinfer1::INetworkDefinition::addUnary	(	ITensor &	input,
		UnaryOperation	operation
	)

pure virtual

Add a unary layer to the network.

Parameters

input	the input tensor to the layer
operation	the operation to apply

See Also: IUnaryLayer

Returns: the new unary layer, or null if it could not be created

virtual IOutputDimensionsFormula& nvinfer1::INetworkDefinition::getConvolutionOutputDimensionsFormula ( ) const

pure virtual

get the convolution output dimensions formula

Deprecated:: this method does not currently work reliably and will be removed in a future release

Returns: the formula from computing the convolution output dimensions

See Also: IOutputDimensionsFormula setConvolutionOutputDimensionsFormula()

virtual IOutputDimensionsFormula& nvinfer1::INetworkDefinition::getDeconvolutionOutputDimensionsFormula ( ) const

pure virtual

get the deconvolution output dimensions formula

Returns: the formula from computing the deconvolution output dimensions.

Deprecated:: this method does not currently work reliably and will be removed in a future release

See Also: IOutputDimensionsFormula setDeconvolutionOutputDimensionsFormula()

virtual ITensor* nvinfer1::INetworkDefinition::getInput ( int index ) const

pure virtual

get the input tensor specified by the given index

Parameters

index the index of the input tensor

Returns: the input tensor, or null if the index is out of range

See Also: getNbInputs()

virtual ILayer* nvinfer1::INetworkDefinition::getLayer ( int index ) const

pure virtual

get the layer specified by the given index

Parameters

index the index of the layer

Returns: the layer, or null if the index is out of range

See Also: getNbLayers()

virtual int nvinfer1::INetworkDefinition::getNbInputs ( ) const

pure virtual

get the number of inputs in the network

Returns: the number of inputs in the network

See Also: getInput()

virtual int nvinfer1::INetworkDefinition::getNbLayers ( ) const

pure virtual

get the number of layers in the network

Returns: the number of layers in the network

See Also: getLayer()

virtual int nvinfer1::INetworkDefinition::getNbOutputs ( ) const

pure virtual

get the number of outputs in the network

Returns: the number of outputs in the network

See Also: getOutput()

virtual ITensor* nvinfer1::INetworkDefinition::getOutput ( int index ) const

pure virtual

get the output tensor specified by the given index

Parameters

index the index of the output tensor

Returns: the output tensor, or null if the index is out of range

See Also: getNbOutputs()

virtual IOutputDimensionsFormula& nvinfer1::INetworkDefinition::getPoolingOutputDimensionsFormula ( ) const

pure virtual

get the pooling output dimensions formula

Returns: the formula from computing the pooling output dimensions

See Also: IOutputDimensionsFormula setPoolingOutputDimensionsFormula()

virtual void nvinfer1::INetworkDefinition::markOutput ( ITensor & tensor )

pure virtual

mark a tensor as a network output

Parameters

tensor the tensor to mark as an output tensor

virtual void nvinfer1::INetworkDefinition::setConvolutionOutputDimensionsFormula ( IOutputDimensionsFormula * formula )

pure virtual

set the convolution output dimensions formula

Deprecated:: this method does not currently work reliably and will be removed in a future release

Parameters

formula the formula from computing the convolution output dimensions. If null is passed, the default formula is used.

the default formula in each dimension is (inputDim + padding * 2 - kernelSize) / stride + 1

See Also: IOutputDimensionsFormula getConvolutionOutputDimensionsFormula()

virtual void nvinfer1::INetworkDefinition::setDeconvolutionOutputDimensionsFormula ( IOutputDimensionsFormula * formula )

pure virtual

set the deconvolution output dimensions formula

Deprecated:: this method does not currently work reliably and will be removed in a future release

Parameters

formula the formula from computing the deconvolution output dimensions. If null is passed, the default formula is used.

the default formula in each dimension is (inputDim - 1) * stride + kernelSize - 2 * padding

See Also: IOutputDimensionsFormula getDevonvolutionOutputDimensionsFormula()

virtual void nvinfer1::INetworkDefinition::setPoolingOutputDimensionsFormula ( IOutputDimensionsFormula * formula )

pure virtual

set the pooling output dimensions formula

Parameters

formula the formula from computing the pooling output dimensions. If null is passed, the default formula is used.

the default formula in each dimension is (inputDim + padding * 2 - kernelSize) / stride + 1

See Also: IOutputDimensionsFormula getPoolingOutputDimensionsFormula()

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

NvInfer.h

Public Member Functions

Detailed Description

Member Function Documentation