Public Member Functions
	Tensor (std::initializer_list< std::int64_t > dims)
	Construct a tensor with the given `dims`.

	Tensor (std::int64_t value)
	Construct a tensor with the given `value`.

	Tensor (float value)
	Construct a tensor with the given `value`.

	Tensor (const std::vector< std::int64_t > &dims)
	Construct a tensor with the given `dims`.

	Tensor (xt::xarray< Expression > values)
	Construct a tensor with the given `values`.

	Tensor (const ::onnx::TensorProto &tensor)
	Construct a tensor with the given `tensor`.

	Tensor (const ::onnx::ValueInfoProto &value_info, const std::string &name)
	Construct a tensor with the given `value_info`.

const xt::xarray< Expression > &	values () const
	Get read-only access to the values of the tensor.

std::size_t	size () const
	Get read-only access to the size of the tensor.

std::size_t	ndim () const
	Get read-only access to the number of dimensions of the tensor.

std::vector< std::int64_t >	dims () const
	Get read-only access to the dimensions of the tensor.

std::int64_t	dim (std::size_t i) const
	Get the dimension at index `i` of the tensor.

bool	SameDim (const Tensor &o) const
	Check whether the two tensors have the same dimension.

bool	Equal (const Tensor &o) const
	Compare two tensor to determine if they are equal.

Tensor &	Flatten (std::int64_t axis)
	Flatten the tensor along the given `axis`.

Tensor &	Transpose (const std::vector< std::int64_t > &perm={})
	Transpose the tensor with the given `perm`.

Tensor &	Reshape (std::initializer_list< std::int64_t > dims)
	Reshape the tensor with the given `dims`.

Tensor &	Reshape (const Tensor &tensor_dim, bool allow_zero)
	Reshape the tensor with the given `tensor_dim`.

Tensor &	Unsqueeze (const Tensor &axes)
	Insert single-dimensional entries to the shape of an input tensor.

Tensor &	Abs ()
	Apply the Abs function to the tensor.

Tensor &	Elementwise (const std::function< Expression(Expression)> &f)
	Apply the `f` function to each element of the tensor.

Tensor &	Slice (const std::vector< std::int64_t > &starts, const std::vector< std::int64_t > &ends, const std::vector< std::int64_t > &axes={}, const std::vector< std::int64_t > &steps={})
	Slice the tensor with the given `starts`, `ends`, `axes`, and `steps`.

Tensor &	Slice (const Tensor &starts, const Tensor &ends, const Tensor &axes={}, const Tensor &steps={})
	Slice the tensor with the given `starts`, `ends`, `axes`, and `steps`.

Tensor	Concat (const Tensor &rhs, std::int64_t axis)
	Concatenate the tensor with the given `rhs` along the given `axis`.

Tensor	Concat (const std::vector< Tensor > &rhs, std::int64_t axis)
	Concatenate the tensors with the given `rhs` along the given `axis`.

Tensor	Gather (const Tensor &indices, std::int64_t axis)
	Gather the tensor with the given `indices` along the given `axis`.

Tensor	MatMul (const Tensor &tensor) const
	Matrix multiplication of two tensors.

Tensor	Convolution (const Tensor &w, const std::vector< std::int64_t > &dilations, std::int64_t group, const std::vector< std::int64_t > &kernel_shape, const std::vector< std::int64_t > &pads, const std::vector< std::int64_t > &strides) const
	Convolution of two tensors.

xt::xarray< Expression >	Convolution (const ImageView &image, const KernelView &kernel, const std::vector< std::size_t > &new_shape, const std::vector< std::int64_t > &dilations, std::int64_t group, const std::vector< std::int64_t > &pads, const std::vector< std::int64_t > &strides) const
	Convolution of two tensors.

Tensor	Pad (const std::vector< std::int64_t > &pads) const
	Pad the tensor with the given `pads`.

Private Member Functions
std::size_t	ComputeOffset (std::initializer_list< std::int64_t > dims) const
	Given a set of indices `dims`, compute the offset of the tensor.

std::size_t	ComputeOffset (const std::int64_t *dims, std::size_t size) const
	Given a set of indices `dims`, compute the offset of the tensor.

Private Attributes
xt::xarray< Expression >	values_
	Internal storage of the values of the tensor.

Detailed Description

Definition at line 22 of file Tensor.h.

Constructor & Destructor Documentation

◆ Tensor() [1/7]

dlinear::onnx::Tensor::Tensor ( std::initializer_list< std::int64_t > dims )

Construct a tensor with the given dims.

Parameters

dims	dimensions of the tensor

Definition at line 56 of file Tensor.cpp.

◆ Tensor() [2/7]

dlinear::onnx::Tensor::Tensor ( std::int64_t value )

explicit

Construct a tensor with the given value.

Parameters

value single value of the tensor

Definition at line 58 of file Tensor.cpp.

◆ Tensor() [3/7]

dlinear::onnx::Tensor::Tensor ( float value )

explicit

Construct a tensor with the given value.

Parameters

value single value of the tensor

Definition at line 60 of file Tensor.cpp.

◆ Tensor() [4/7]

dlinear::onnx::Tensor::Tensor ( const std::vector< std::int64_t > & dims )

explicit

Construct a tensor with the given dims.

Parameters

dims	dimensions of the tensor

Definition at line 62 of file Tensor.cpp.

◆ Tensor() [5/7]

dlinear::onnx::Tensor::Tensor ( xt::xarray< Expression > values )

explicit

Construct a tensor with the given values.

Parameters

values xarray of expressions

Definition at line 64 of file Tensor.cpp.

◆ Tensor() [6/7]

dlinear::onnx::Tensor::Tensor ( const ::onnx::TensorProto & tensor )

explicit

Construct a tensor with the given tensor.

Parameters

tensor tensor proto from the ONNX model

Definition at line 72 of file Tensor.cpp.

◆ Tensor() [7/7]

dlinear::onnx::Tensor::Tensor	(	const ::onnx::ValueInfoProto &	value_info,
		const std::string &	name )

explicit

Construct a tensor with the given value_info.

Parameters

value_info	value info proto from the ONNX model
name	prefix name to give to the variables in the tensor

Definition at line 66 of file Tensor.cpp.

Member Function Documentation

◆ Abs()

Tensor & dlinear::onnx::Tensor::Abs ( )

Apply the Abs function to the tensor.

ONNX documentation.

Returns: reference to the tensor

Definition at line 216 of file Tensor.cpp.

◆ ComputeOffset() [1/2]

std::size_t dlinear::onnx::Tensor::ComputeOffset	(	const std::int64_t *	dims,
		std::size_t	size ) const

nodiscardprivate

Given a set of indices dims, compute the offset of the tensor.

Parameters

dims	set of indices
size	size of the set of indices

Returns: offset of the tensor

Definition at line 603 of file Tensor.cpp.

◆ ComputeOffset() [2/2]

std::size_t dlinear::onnx::Tensor::ComputeOffset ( std::initializer_list< std::int64_t > dims ) const

nodiscardprivate

Given a set of indices dims, compute the offset of the tensor.

Parameters

dims	set of indices

Returns: offset of the tensor

Definition at line 599 of file Tensor.cpp.

◆ Concat() [1/2]

Tensor dlinear::onnx::Tensor::Concat	(	const std::vector< Tensor > &	rhs,
		std::int64_t	axis )

Concatenate the tensors with the given rhs along the given axis.

ONNX documentation.

Parameters

rhs	other tensors to concatenate with
axis	Which axis to concat on. A negative value means counting dimensions from the back Accepted range is \([-r, r-1]\) where \(r = \text{rank(inputs)}\)

Returns: new tensor with the concatenated values

Definition at line 258 of file Tensor.cpp.

◆ Concat() [2/2]

Tensor dlinear::onnx::Tensor::Concat	(	const Tensor &	rhs,
		std::int64_t	axis )

Concatenate the tensor with the given rhs along the given axis.

ONNX documentation.

Parameters

rhs	other tensor to concatenate with
axis	Which axis to concat on. A negative value means counting dimensions from the back Accepted range is \([-r, r-1]\) where \(r = \text{rank(inputs)}\)

Returns: new tensor with the concatenated values

Definition at line 254 of file Tensor.cpp.

◆ Convolution() [1/2]

xt::xarray< Expression > dlinear::onnx::Tensor::Convolution	(	const ImageView &	image,
		const KernelView &	kernel,
		const std::vector< std::size_t > &	new_shape,
		const std::vector< std::int64_t > &	dilations,
		std::int64_t	group,
		const std::vector< std::int64_t > &	pads,
		const std::vector< std::int64_t > &	strides ) const

nodiscard

Convolution of two tensors.

ONNX documentation.

Parameters

image	input tensor
kernel	convolution kernel
new_shape	shape of the output tensor
dilations	dilation value along each spatial axis of the filter
group	number of groups input channels and output channels are divided into
pads	padding for the beginning and ending along each spatial axis, it can take any value greater than or equal to 0. The value represent the number of pixels added to the beginning and end part of the corresponding axis. pads format should be as follow \([x1_{begin}, x2_{begin} \dots x1_{end}, x2_{end} \dots]\), where \(xi_{begin}\) is the number of pixels added at the beginning of axis \(i\) and \(xi_{end}\) is the number of pixels added at the end of axis \(i\).
strides	stride along each spatial axis

Returns

Definition at line 352 of file Tensor.cpp.

◆ Convolution() [2/2]

Tensor dlinear::onnx::Tensor::Convolution	(	const Tensor &	w,
		const std::vector< std::int64_t > &	dilations,
		std::int64_t	group,
		const std::vector< std::int64_t > &	kernel_shape,
		const std::vector< std::int64_t > &	pads,
		const std::vector< std::int64_t > &	strides ) const

nodiscard

Convolution of two tensors.

ONNX documentation.

Parameters

w	convolution kernel with shape \([M, C, k_1, k_2, \dots, k_n]\)
dilations	dilation value along each spatial axis of the filter
group	number of groups input channels and output channels are divided into
kernel_shape	shape of the convolution kernel. If not present, should be inferred from input W
pads	padding for the beginning and ending along each spatial axis, it can take any value greater than or equal to 0. The value represent the number of pixels added to the beginning and end part of the corresponding axis. pads format should be as follow \([x1_{begin}, x2_{begin} \dots x1_{end}, x2_{end} \dots]\), where \(xi_{begin}\) is the number of pixels added at the beginning of axis \(i\) and \(xi_{end}\) is the number of pixels added at the end of axis \(i\).
strides	stride along each spatial axis

Returns

Definition at line 302 of file Tensor.cpp.

◆ dim()

std::int64_t dlinear::onnx::Tensor::dim ( std::size_t i ) const

nodiscard

Get the dimension at index i of the tensor.

If the dimension requested is not present, it will return 1.

Parameters

i	index of the dimension

Returns: the dimension at the given index; 1 if the i is out of bounds

Definition at line 142 of file Tensor.cpp.

◆ dims()

std::vector< std::int64_t > dlinear::onnx::Tensor::dims ( ) const

nodiscard

Get read-only access to the dimensions of the tensor.

Returns: dimensions of the tensor

Definition at line 140 of file Tensor.cpp.

◆ Elementwise()

Tensor & dlinear::onnx::Tensor::Elementwise ( const std::function< Expression(Expression)> & f )

Apply the f function to each element of the tensor.

Parameters

f	function to apply to each element

Returns: reference to the tensor

Definition at line 221 of file Tensor.cpp.

◆ Equal()

bool dlinear::onnx::Tensor::Equal ( const Tensor & o ) const

nodiscard

Compare two tensor to determine if they are equal.

Parameters

o	other object to compare against

Returns: true the two object are equal; false if the two objects are different

Definition at line 152 of file Tensor.cpp.

◆ Flatten()

Tensor & dlinear::onnx::Tensor::Flatten ( std::int64_t axis )

Flatten the tensor along the given axis.

ONNX documentation.

Parameters

axis	axis to flatten along

Returns: reference to the tensor

Definition at line 160 of file Tensor.cpp.

◆ Gather()

Tensor dlinear::onnx::Tensor::Gather	(	const Tensor &	indices,
		std::int64_t	axis )

Gather the tensor with the given indices along the given axis.

ONNX documentation.

Parameters

indices	tensor of any rank \(q\). All index values are expected to be within bounds \([-s, s-1]\) along axis of size \(s\). It is an error if any of the index values are out of bounds
axis	which axis to concat on. A negative value means counting dimensions from the back Accepted range is \([-r, r-1]\) where \(r = \text{rank(inputs)}\)

Returns: new tensor with the gathered values

Definition at line 265 of file Tensor.cpp.

◆ MatMul()

Tensor dlinear::onnx::Tensor::MatMul ( const Tensor & tensor ) const

nodiscard

Matrix multiplication of two tensors.

ONNX documentation.

Parameters

tensor tensor to multiply with

Returns: new tensor with the multiplied values

Definition at line 436 of file Tensor.cpp.

◆ ndim()

std::size_t dlinear::onnx::Tensor::ndim ( ) const

inlinenodiscard

Get read-only access to the number of dimensions of the tensor.

Returns: number of dimensions of the tensor

Definition at line 69 of file Tensor.h.

◆ Pad()

Tensor dlinear::onnx::Tensor::Pad ( const std::vector< std::int64_t > & pads ) const

nodiscard

Pad the tensor with the given pads.

ONNX documentation.

Parameters

pads	padding for the beginning and ending along each spatial axis

Definition at line 412 of file Tensor.cpp.

◆ Reshape() [1/2]

Tensor & dlinear::onnx::Tensor::Reshape	(	const Tensor &	tensor_dim,
		bool	allow_zero )

Reshape the tensor with the given tensor_dim.

ONNX documentation.

Parameters

tensor_dim	tensor containing the dimensions to reshape the tensor
allow_zero	whether to allow zero in the dimensions

Returns: reference to the tensor

Definition at line 185 of file Tensor.cpp.

◆ Reshape() [2/2]

Tensor & dlinear::onnx::Tensor::Reshape ( std::initializer_list< std::int64_t > dims )

Reshape the tensor with the given dims.

ONNX documentation.

Parameters

dims	dimensions to reshape the tensor

Returns: reference to the tensor

Definition at line 181 of file Tensor.cpp.

◆ SameDim()

bool dlinear::onnx::Tensor::SameDim ( const Tensor & o ) const

nodiscard

Check whether the two tensors have the same dimension.

Parameters

o	tensor to compare with

Returns: true if the two tensors have the same dimension; false if the two tensors have different dimensions

Definition at line 147 of file Tensor.cpp.

◆ size()

std::size_t dlinear::onnx::Tensor::size ( ) const

inlinenodiscard

Get read-only access to the size of the tensor.

Returns: size of the tensor

Definition at line 67 of file Tensor.h.

◆ Slice() [1/2]

Tensor & dlinear::onnx::Tensor::Slice	(	const std::vector< std::int64_t > &	starts,
		const std::vector< std::int64_t > &	ends,
		const std::vector< std::int64_t > &	axes = {},
		const std::vector< std::int64_t > &	steps = {} )

Slice the tensor with the given starts, ends, axes, and steps.

ONNX documentation.

Parameters

starts	1-D tensor of starting indices of corresponding axis in `axes`
ends	1-D tensor of ending indices (exclusive) of corresponding axis in `axes`
axes	1-D tensor of axes that `starts` and `ends` apply to. Negative value means counting dimensions from the back. Accepted range is \([-r, r-1]\) where \(r = \text{rank(input)}\). Behavior is undefined if an axis is repeated.
steps	1-D tensor of slice step of corresponding axis in `axes`. Negative value means slicing backward. Cannot be 0. Defaults to 1.

Returns: reference to the tensor

Definition at line 230 of file Tensor.cpp.

◆ Slice() [2/2]

Tensor & dlinear::onnx::Tensor::Slice	(	const Tensor &	starts,
		const Tensor &	ends,
		const Tensor &	axes = {},
		const Tensor &	steps = {} )

Slice the tensor with the given starts, ends, axes, and steps.

ONNX documentation.

Parameters

starts	1-D tensor of starting indices of corresponding axis in `axes`
ends	1-D tensor of ending indices (exclusive) of corresponding axis in `axes`
axes	1-D tensor of axes that `starts` and `ends` apply to. Negative value means counting dimensions from the back. Accepted range is \([-r, r-1]\) where \(r = \text{rank(input)}\). Behavior is undefined if an axis is repeated
steps	1-D tensor of slice step of corresponding axis in `axes`. Negative value means slicing backward. Cannot be 0. Defaults to 1

Returns: reference to the tensor

Definition at line 226 of file Tensor.cpp.

◆ Transpose()

Tensor & dlinear::onnx::Tensor::Transpose ( const std::vector< std::int64_t > & perm = {} )

Transpose the tensor with the given perm.

ONNX documentation.

Parameters

perm	permutation of the dimensions

Returns: reference to the tensor

Definition at line 211 of file Tensor.cpp.

◆ Unsqueeze()

Tensor & dlinear::onnx::Tensor::Unsqueeze ( const Tensor & axes )

Insert single-dimensional entries to the shape of an input tensor.

ONNX documentation.

Parameters

axes	axes along which the tensor will be unsqueezed

Returns: reference to the tensor

Definition at line 171 of file Tensor.cpp.

◆ values()

const xt::xarray< Expression > & dlinear::onnx::Tensor::values ( ) const

inlinenodiscard

Get read-only access to the values of the tensor.

Returns: values of the tensor

Definition at line 65 of file Tensor.h.

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

dlinear/parser/onnx/Tensor.h
dlinear/parser/onnx/Tensor.cpp

Public Member Functions

Private Member Functions

Private Attributes

Detailed Description

Constructor & Destructor Documentation

◆ Tensor() [1/7]

◆ Tensor() [2/7]

◆ Tensor() [3/7]

◆ Tensor() [4/7]

◆ Tensor() [5/7]

◆ Tensor() [6/7]

◆ Tensor() [7/7]

Member Function Documentation

◆ Abs()

◆ ComputeOffset() [1/2]

◆ ComputeOffset() [2/2]

◆ Concat() [1/2]

◆ Concat() [2/2]

◆ Convolution() [1/2]

◆ Convolution() [2/2]

◆ dim()

◆ dims()

◆ Elementwise()

◆ Equal()

◆ Flatten()

◆ Gather()

◆ MatMul()

◆ ndim()

◆ Pad()

◆ Reshape() [1/2]

◆ Reshape() [2/2]

◆ SameDim()

◆ size()

◆ Slice() [1/2]

◆ Slice() [2/2]

◆ Transpose()

◆ Unsqueeze()

◆ values()