Device-aware N-dimensional tensor. More...

Inheritance diagram for Mila::Dnn::Tensor< TDataType, TMemoryResource >:

Collaboration diagram for Mila::Dnn::Tensor< TDataType, TMemoryResource >:

[legend]

Public Types
using	DataType = TensorDataType
	Abstract data type enumeration.
using	DataTypeTraits = TensorDataTypeTraits<TDataType>
	Compile-time data type characteristics.
using	host_value_t = std::conditional_t<TensorDataTypeTraits<TDataType>::is_integer_type, int32_t, float>
	Host value type for scalars.
using	MemoryResource = TMemoryResource
	Memory resource type for this tensor.

Public Member Functions
	Tensor (Compute::DeviceId device_id, const shape_t &shape, std::string name={})
	Creates a tensor bound to the specified device id and shape.
	Tensor (const Tensor &other)=delete
	Copy constructor - explicitly deleted for performance safety.
	Tensor (Tensor &&other) noexcept
	Efficiently transfers ownership from another tensor.
	~Tensor ()=default
	Destructor with automatic resource cleanup via RAII.
constexpr const auto *	data () const noexcept
	Returns type-safe immutable pointer to tensor data with concrete host type.
constexpr auto *	data () noexcept
	Returns type-safe pointer to tensor data with concrete host type.
size_t	elementSize () const override
	Returns the size in bytes of a single tensor element.
bool	empty () const
	Checks if the tensor contains no elements.
Tensor< TDataType, TMemoryResource > &	flatten ()
	Flattens tensor to 2D shape in-place.
TensorDataType	getDataType () const override
	Returns the tensor's abstract data type identifier.
std::string	getDataTypeName () const override
	Returns human-readable name of the tensor's data type.
Compute::DeviceId	getDeviceId () const override
Compute::DeviceType	getDeviceType () const override
	Returns the device type of this tensor's memory resource.
Compute::MemoryResource *	getMemoryResource () const override
	Returns pointer to the memory resource managing this tensor's storage.
std::string	getName () const override
	Returns the tensor's optional user-assigned name.
size_t	getStorageSize () const override
	Returns the total storage size in bytes backing the tensor's buffer.
std::string	getUId () const override
	Returns the tensor's unique identifier.
bool	isScalar () const noexcept override
	Checks if tensor is a scalar (0-dimensional).
bool	isValid () const noexcept
	Check if tensor is in a valid state (not moved-from).
bool	isView () const override
	Check if this tensor is a view of another tensor's data.
auto &	item ()
	Gets the scalar value for 0-dimensional tensors.
const auto &	item () const
	Gets the scalar value for 0-dimensional tensors (const version).
Tensor &	operator= (const Tensor &other)=delete
	Copy assignment operator - explicitly deleted.
Tensor &	operator= (Tensor &&other) noexcept
	Efficiently moves resources from another tensor.
auto &	operator[] (const index_t &indices)
	Accesses tensor element using multi-dimensional indices.
const auto &	operator[] (const index_t &indices) const
	Accesses tensor element using multi-dimensional indices (const version).
template<typename... Indices> requires TMemoryResource::is_host_accessible
auto &	operator[] (Indices... indices) &&(std
	Variadic index access (non-const).
template<typename... Indices> requires TMemoryResource::is_host_accessible
const auto &	operator[] (Indices... indices) const &&(std
	Variadic index access (const).
size_t	rank () const
	Returns the number of dimensions in the tensor.
const void *	rawData () const override
	Returns raw pointer to tensor data (const version).
void *	rawData () override
	Returns raw pointer to tensor data (implements ITensor protected API).
void	reshape (const shape_t &new_shape)
	Modifies tensor shape while preserving total element count.
void	setName (const std::string &value)
	Assigns a descriptive name to the tensor.
const shape_t &	shape () const override
	Returns the tensor's dimensional shape vector.
size_t	size () const override
	Returns the total number of logical elements in the tensor.
const stride_t &	strides () const
	Returns the tensor's memory stride information.
std::string	toString (bool showBuffer=false) const
	Generates string representation of the tensor.
Tensor	view (const shape_t &new_shape, size_t offset=0) const
Public Member Functions inherited from Mila::Dnn::ITensor
virtual	~ITensor ()=default
template<typename T>
bool	isType () const
	Type-safe check for tensor element type.

Static Public Member Functions
static constexpr bool	is_device_accessible ()
	Checks if tensor data is accessible from device (GPU) code.
static constexpr bool	is_host_accessible ()
	Checks if tensor data is accessible from host (CPU) code.

Private Member Functions
	Tensor (const Tensor &parent, const shape_t &shape, size_t offset)
void	allocateBuffer ()
	Allocates and initializes the tensor's data buffer.
int64_t	computeFlatIndex (const index_t &indices) const
	Computes flat memory index from multi-dimensional indices.
int64_t	computeSize (const shape_t &shape) const
	Computes total element count from shape vector.
stride_t	computeStrides (const shape_t &shape)
	Computes row-major memory strides from shape.
std::string	getBufferString () const
	Gets formatted buffer content for string representation.
std::string	outputBuffer (size_t index, size_t depth) const
std::string	outputLayout () const
	Generates formatted layout information string.
std::string	setUId ()
	Generates unique identifier for new tensor instance.
void	validateIndices (const index_t &indices, const char *fn) const
	Validates multi-dimensional indices against tensor shape.

Static Private Member Functions
static Compute::DeviceId	validateDeviceId (Compute::DeviceId device_id)
	Validates device id against the memory resource requirement.

Private Attributes
std::shared_ptr< TensorBuffer< TDataType, TMemoryResource > >	buffer_ { nullptr }
	Managed buffer containing tensor data.
Compute::DeviceId	device_id_
	DeviceId for proper device binding and resource management.
bool	is_view_ { false }
std::string	name_
	Optional user-assigned name for debugging.
shape_t	shape_ {}
	Dimensional sizes for each tensor dimension.
size_t	size_ { 0 }
	Total number of logical elements in the tensor.
stride_t	strides_ {}
	Memory stride values for multi-dimensional indexing.
std::string	uid_
	Unique identifier for this tensor instance.
size_t	view_offset_ { 0 }

Detailed Description

template<TensorDataType TDataType, typename TMemoryResource>
requires isValidTensor<TDataType, TMemoryResource>
class Mila::Dnn::Tensor< TDataType, TMemoryResource >

Device-aware N-dimensional tensor.

Move-only tensor parameterized by an abstract TensorDataType and a MemoryResource. Scalars are represented by an empty shape ({}), which yields size() == 1.

Template Parameters

TDataType	Abstract tensor data type from TensorDataType enumeration
TMemoryResource	Memory resource type defining storage location and access patterns

Examples: /__w/Mila/Mila/Mila/Src/Dnn/Components/Activations/Gelu/Gelu.ixx, and /__w/Mila/Mila/Mila/Src/Dnn/Components/Normalization/Softmax.ixx.

Member Typedef Documentation

◆ DataType

template<TensorDataType TDataType, typename TMemoryResource>

using Mila::Dnn::Tensor< TDataType, TMemoryResource >::DataType = TensorDataType

Abstract data type enumeration.

◆ DataTypeTraits

template<TensorDataType TDataType, typename TMemoryResource>

using Mila::Dnn::Tensor< TDataType, TMemoryResource >::DataTypeTraits = TensorDataTypeTraits<TDataType>

Compile-time data type characteristics.

◆ host_value_t

template<TensorDataType TDataType, typename TMemoryResource>

using Mila::Dnn::Tensor< TDataType, TMemoryResource >::host_value_t = std::conditional_t<TensorDataTypeTraits<TDataType>::is_integer_type, int32_t, float>

Host value type for scalars.

◆ MemoryResource

template<TensorDataType TDataType, typename TMemoryResource>

using Mila::Dnn::Tensor< TDataType, TMemoryResource >::MemoryResource = TMemoryResource

Memory resource type for this tensor.

Constructor & Destructor Documentation

◆ Tensor() [1/4]

template<TensorDataType TDataType, typename TMemoryResource>

Mila::Dnn::Tensor< TDataType, TMemoryResource >::Tensor	(	Compute::DeviceId	device_id,
		const shape_t &	shape,
		std::string	name = {} )

inlineexplicit

Creates a tensor bound to the specified device id and shape.

Constructs a tensor using a DeviceId. Empty shape {} creates a scalar (0D tensor) with size 1. Shape containing a zero produces an empty tensor.

Parameters

device_id	Device identifier (type + index)
shape	Vector defining the size of each dimension in row-major order

Exceptions

std::invalid_argument	If device id is invalid (checked by validateDeviceId)
std::runtime_error	If device type doesn't match memory resource
std::bad_alloc	If memory allocation fails

Here is the caller graph for this function:

◆ Tensor() [2/4]

template<TensorDataType TDataType, typename TMemoryResource>

Mila::Dnn::Tensor< TDataType, TMemoryResource >::Tensor ( const Tensor< TDataType, TMemoryResource > & other )

delete

Copy constructor - explicitly deleted for performance safety.

Use clone() for deep copies or std::move() for ownership transfer.

◆ Tensor() [3/4]

template<TensorDataType TDataType, typename TMemoryResource>

Mila::Dnn::Tensor< TDataType, TMemoryResource >::Tensor ( Tensor< TDataType, TMemoryResource > && other )

inlinenoexcept

Efficiently transfers ownership from another tensor.

Moves resources from the source tensor, leaving it in a moved-from state.

◆ ~Tensor()

template<TensorDataType TDataType, typename TMemoryResource>

Mila::Dnn::Tensor< TDataType, TMemoryResource >::~Tensor ( )

default

Destructor with automatic resource cleanup via RAII.

◆ Tensor() [4/4]

template<TensorDataType TDataType, typename TMemoryResource>

Mila::Dnn::Tensor< TDataType, TMemoryResource >::Tensor	(	const Tensor< TDataType, TMemoryResource > &	parent,
		const shape_t &	shape,
		size_t	offset )

inlineprivate

Member Function Documentation

◆ allocateBuffer()

template<TensorDataType TDataType, typename TMemoryResource>

void Mila::Dnn::Tensor< TDataType, TMemoryResource >::allocateBuffer ( )

inlineprivate

Allocates and initializes the tensor's data buffer.

Scalars allocate normally (size==1); truly empty tensors allocate nothing.

Here is the caller graph for this function:

◆ computeFlatIndex()

template<TensorDataType TDataType, typename TMemoryResource>

int64_t Mila::Dnn::Tensor< TDataType, TMemoryResource >::computeFlatIndex ( const index_t & indices ) const

inlineprivate

Computes flat memory index from multi-dimensional indices.

Here is the caller graph for this function:

◆ computeSize()

template<TensorDataType TDataType, typename TMemoryResource>

int64_t Mila::Dnn::Tensor< TDataType, TMemoryResource >::computeSize ( const shape_t & shape ) const

inlineprivate

Computes total element count from shape vector.

Here is the caller graph for this function:

◆ computeStrides()

template<TensorDataType TDataType, typename TMemoryResource>

stride_t Mila::Dnn::Tensor< TDataType, TMemoryResource >::computeStrides ( const shape_t & shape )

inlineprivate

Computes row-major memory strides from shape.

Here is the caller graph for this function:

◆ data() [1/2]

template<TensorDataType TDataType, typename TMemoryResource>

const auto * Mila::Dnn::Tensor< TDataType, TMemoryResource >::data ( ) const

inlinenodiscardconstexprnoexcept

Returns type-safe immutable pointer to tensor data with concrete host type.

◆ data() [2/2]

template<TensorDataType TDataType, typename TMemoryResource>

auto * Mila::Dnn::Tensor< TDataType, TMemoryResource >::data ( )

inlinenodiscardconstexprnoexcept

Returns type-safe pointer to tensor data with concrete host type.

Only available for host-accessible memory resources.

Here is the caller graph for this function:

◆ elementSize()

template<TensorDataType TDataType, typename TMemoryResource>

size_t Mila::Dnn::Tensor< TDataType, TMemoryResource >::elementSize ( ) const

inlineoverridevirtual

Returns the size in bytes of a single tensor element.

Implements Mila::Dnn::ITensor.

Here is the caller graph for this function:

◆ empty()

template<TensorDataType TDataType, typename TMemoryResource>

bool Mila::Dnn::Tensor< TDataType, TMemoryResource >::empty ( ) const

inline

Checks if the tensor contains no elements.

Scalars are NOT empty (size == 1). Empty tensors have size == 0.

Here is the caller graph for this function:

◆ flatten()

template<TensorDataType TDataType, typename TMemoryResource>

Tensor< TDataType, TMemoryResource > & Mila::Dnn::Tensor< TDataType, TMemoryResource >::flatten ( )

inline

Flattens tensor to 2D shape in-place.

No-op for scalars and vectors.

◆ getBufferString()

template<TensorDataType TDataType, typename TMemoryResource>

std::string Mila::Dnn::Tensor< TDataType, TMemoryResource >::getBufferString ( ) const

inlineprivate

Gets formatted buffer content for string representation.

Here is the caller graph for this function:

◆ getDataType()

template<TensorDataType TDataType, typename TMemoryResource>

TensorDataType Mila::Dnn::Tensor< TDataType, TMemoryResource >::getDataType ( ) const

inlineoverridevirtual

Returns the tensor's abstract data type identifier.

Implements Mila::Dnn::ITensor.

◆ getDataTypeName()

template<TensorDataType TDataType, typename TMemoryResource>

std::string Mila::Dnn::Tensor< TDataType, TMemoryResource >::getDataTypeName ( ) const

inlineoverridevirtual

Returns human-readable name of the tensor's data type.

Implements Mila::Dnn::ITensor.

◆ getDeviceId()

template<TensorDataType TDataType, typename TMemoryResource>

Compute::DeviceId Mila::Dnn::Tensor< TDataType, TMemoryResource >::getDeviceId ( ) const

inlineoverridevirtual

Implements Mila::Dnn::ITensor.

Here is the caller graph for this function:

◆ getDeviceType()

template<TensorDataType TDataType, typename TMemoryResource>

Compute::DeviceType Mila::Dnn::Tensor< TDataType, TMemoryResource >::getDeviceType ( ) const

inlineoverridevirtual

Returns the device type of this tensor's memory resource.

Equivalent to TMemoryResource::device_type.

Implements Mila::Dnn::ITensor.

◆ getMemoryResource()

template<TensorDataType TDataType, typename TMemoryResource>

Compute::MemoryResource * Mila::Dnn::Tensor< TDataType, TMemoryResource >::getMemoryResource ( ) const

inlineoverridevirtual

Returns pointer to the memory resource managing this tensor's storage.

Implements Mila::Dnn::ITensor.

◆ getName()

template<TensorDataType TDataType, typename TMemoryResource>

std::string Mila::Dnn::Tensor< TDataType, TMemoryResource >::getName ( ) const

inlineoverridevirtual

Returns the tensor's optional user-assigned name.

Implements Mila::Dnn::ITensor.

◆ getStorageSize()

template<TensorDataType TDataType, typename TMemoryResource>

size_t Mila::Dnn::Tensor< TDataType, TMemoryResource >::getStorageSize ( ) const

inlineoverridevirtual

Returns the total storage size in bytes backing the tensor's buffer.

Forwards to the authoritative TensorBuffer::storageBytes() value and returns 0 when no buffer is allocated.

Implements Mila::Dnn::ITensor.

Here is the caller graph for this function:

◆ getUId()

template<TensorDataType TDataType, typename TMemoryResource>

std::string Mila::Dnn::Tensor< TDataType, TMemoryResource >::getUId ( ) const

inlineoverridevirtual

Returns the tensor's unique identifier.

Implements Mila::Dnn::ITensor.

◆ is_device_accessible()

template<TensorDataType TDataType, typename TMemoryResource>

constexpr bool Mila::Dnn::Tensor< TDataType, TMemoryResource >::is_device_accessible ( )

inlinestaticconstexpr

Checks if tensor data is accessible from device (GPU) code.

◆ is_host_accessible()

template<TensorDataType TDataType, typename TMemoryResource>

constexpr bool Mila::Dnn::Tensor< TDataType, TMemoryResource >::is_host_accessible ( )

inlinestaticconstexpr

Checks if tensor data is accessible from host (CPU) code.

Here is the caller graph for this function:

◆ isScalar()

template<TensorDataType TDataType, typename TMemoryResource>

bool Mila::Dnn::Tensor< TDataType, TMemoryResource >::isScalar ( ) const

inlineoverridevirtualnoexcept

Checks if tensor is a scalar (0-dimensional).

Scalars: shape {} and size() == 1.

Implements Mila::Dnn::ITensor.

Here is the caller graph for this function:

◆ isValid()

template<TensorDataType TDataType, typename TMemoryResource>

bool Mila::Dnn::Tensor< TDataType, TMemoryResource >::isValid ( ) const

inlinenoexcept

Check if tensor is in a valid state (not moved-from).

◆ isView()

template<TensorDataType TDataType, typename TMemoryResource>

bool Mila::Dnn::Tensor< TDataType, TMemoryResource >::isView ( ) const

inlineoverridevirtual

Check if this tensor is a view of another tensor's data.

Returns: true if this is a view, false if it owns its buffer.

Implements Mila::Dnn::ITensor.

◆ item() [1/2]

template<TensorDataType TDataType, typename TMemoryResource>

auto & Mila::Dnn::Tensor< TDataType, TMemoryResource >::item ( )

inline

Gets the scalar value for 0-dimensional tensors.

Only available for host-accessible memory resources.

Exceptions

std::runtime_error If tensor is not a scalar (rank != 0)

◆ item() [2/2]

template<TensorDataType TDataType, typename TMemoryResource>

const auto & Mila::Dnn::Tensor< TDataType, TMemoryResource >::item ( ) const

inline

Gets the scalar value for 0-dimensional tensors (const version).

Only available for host-accessible memory resources.

Exceptions

std::runtime_error If tensor is not a scalar (rank != 0)

◆ operator=() [1/2]

template<TensorDataType TDataType, typename TMemoryResource>

Tensor & Mila::Dnn::Tensor< TDataType, TMemoryResource >::operator= ( const Tensor< TDataType, TMemoryResource > & other )

delete

Copy assignment operator - explicitly deleted.

Use clone() for deep copies or std::move() for transfers.

◆ operator=() [2/2]

template<TensorDataType TDataType, typename TMemoryResource>

Tensor & Mila::Dnn::Tensor< TDataType, TMemoryResource >::operator= ( Tensor< TDataType, TMemoryResource > && other )

inlinenoexcept

Efficiently moves resources from another tensor.

Self-assignment safe implementation.

◆ operator[]() [1/4]

template<TensorDataType TDataType, typename TMemoryResource>

auto & Mila::Dnn::Tensor< TDataType, TMemoryResource >::operator[] ( const index_t & indices )

inline

Accesses tensor element using multi-dimensional indices.

Only available for host-accessible memory resources.

Here is the caller graph for this function:

◆ operator[]() [2/4]

template<TensorDataType TDataType, typename TMemoryResource>

const auto & Mila::Dnn::Tensor< TDataType, TMemoryResource >::operator[] ( const index_t & indices ) const

inline

Accesses tensor element using multi-dimensional indices (const version).

◆ operator[]() [3/4]

template<TensorDataType TDataType, typename TMemoryResource>

template<typename... Indices>
requires TMemoryResource::is_host_accessible

auto & Mila::Dnn::Tensor< TDataType, TMemoryResource >::operator[] ( Indices... indices ) &&

inline

Variadic index access (non-const).

◆ operator[]() [4/4]

template<TensorDataType TDataType, typename TMemoryResource>

template<typename... Indices>
requires TMemoryResource::is_host_accessible

const auto & Mila::Dnn::Tensor< TDataType, TMemoryResource >::operator[] ( Indices... indices ) const &&

inline

Variadic index access (const).

◆ outputBuffer()

template<TensorDataType TDataType, typename TMemoryResource>

std::string Mila::Dnn::Tensor< TDataType, TMemoryResource >::outputBuffer	(	size_t	index,
		size_t	depth ) const

inlineprivate

Here is the caller graph for this function:

◆ outputLayout()

template<TensorDataType TDataType, typename TMemoryResource>

std::string Mila::Dnn::Tensor< TDataType, TMemoryResource >::outputLayout ( ) const

inlineprivate

Generates formatted layout information string.

Here is the caller graph for this function:

◆ rank()

template<TensorDataType TDataType, typename TMemoryResource>

size_t Mila::Dnn::Tensor< TDataType, TMemoryResource >::rank ( ) const

inline

Returns the number of dimensions in the tensor.

Here is the caller graph for this function:

◆ rawData() [1/2]

template<TensorDataType TDataType, typename TMemoryResource>

const void * Mila::Dnn::Tensor< TDataType, TMemoryResource >::rawData ( ) const

inlineoverridevirtual

Returns raw pointer to tensor data (const version).

Implements Mila::Dnn::ITensor.

◆ rawData() [2/2]

template<TensorDataType TDataType, typename TMemoryResource>

void * Mila::Dnn::Tensor< TDataType, TMemoryResource >::rawData ( )

inlineoverridevirtual

Returns raw pointer to tensor data (implements ITensor protected API).

Implements Mila::Dnn::ITensor.

Here is the caller graph for this function:

◆ reshape()

template<TensorDataType TDataType, typename TMemoryResource>

void Mila::Dnn::Tensor< TDataType, TMemoryResource >::reshape ( const shape_t & new_shape )

inline

Modifies tensor shape while preserving total element count.

The new shape must have the same total number of elements as the current shape, unless the tensor is empty.

Here is the caller graph for this function:

◆ setName()

template<TensorDataType TDataType, typename TMemoryResource>

void Mila::Dnn::Tensor< TDataType, TMemoryResource >::setName ( const std::string & value )

inline

Assigns a descriptive name to the tensor.

Names must be non-empty strings.

Here is the caller graph for this function:

◆ setUId()

template<TensorDataType TDataType, typename TMemoryResource>

std::string Mila::Dnn::Tensor< TDataType, TMemoryResource >::setUId ( )

inlineprivate

Generates unique identifier for new tensor instance.

◆ shape()

template<TensorDataType TDataType, typename TMemoryResource>

const shape_t & Mila::Dnn::Tensor< TDataType, TMemoryResource >::shape ( ) const

inlineoverridevirtual

Returns the tensor's dimensional shape vector.

Implements Mila::Dnn::ITensor.

Here is the caller graph for this function:

◆ size()

template<TensorDataType TDataType, typename TMemoryResource>

size_t Mila::Dnn::Tensor< TDataType, TMemoryResource >::size ( ) const

inlineoverridevirtual

Returns the total number of logical elements in the tensor.

Implements Mila::Dnn::ITensor.

Here is the caller graph for this function:

◆ strides()

template<TensorDataType TDataType, typename TMemoryResource>

const stride_t & Mila::Dnn::Tensor< TDataType, TMemoryResource >::strides ( ) const

inline

Returns the tensor's memory stride information.

◆ toString()

template<TensorDataType TDataType, typename TMemoryResource>

std::string Mila::Dnn::Tensor< TDataType, TMemoryResource >::toString ( bool showBuffer = false ) const

inline

Generates string representation of the tensor.

Includes UID, name, shape, data type and device information.

Here is the caller graph for this function:

◆ validateDeviceId()

template<TensorDataType TDataType, typename TMemoryResource>

Compute::DeviceId Mila::Dnn::Tensor< TDataType, TMemoryResource >::validateDeviceId ( Compute::DeviceId device_id )

inlinestaticprivate

Validates device id against the memory resource requirement.

Throws if types do not match.

◆ validateIndices()

template<TensorDataType TDataType, typename TMemoryResource>

void Mila::Dnn::Tensor< TDataType, TMemoryResource >::validateIndices	(	const index_t &	indices,
		const char *	fn ) const

inlineprivate

Validates multi-dimensional indices against tensor shape.

Here is the caller graph for this function:

◆ view()

template<TensorDataType TDataType, typename TMemoryResource>

Tensor Mila::Dnn::Tensor< TDataType, TMemoryResource >::view	(	const shape_t &	new_shape,
		size_t	offset = 0 ) const

inlinenodiscard

Here is the caller graph for this function:

Member Data Documentation

◆ buffer_

template<TensorDataType TDataType, typename TMemoryResource>

std::shared_ptr<TensorBuffer<TDataType, TMemoryResource> > Mila::Dnn::Tensor< TDataType, TMemoryResource >::buffer_ { nullptr }

private

Managed buffer containing tensor data.

◆ device_id_

template<TensorDataType TDataType, typename TMemoryResource>

Compute::DeviceId Mila::Dnn::Tensor< TDataType, TMemoryResource >::device_id_

private

DeviceId for proper device binding and resource management.

◆ is_view_

template<TensorDataType TDataType, typename TMemoryResource>

bool Mila::Dnn::Tensor< TDataType, TMemoryResource >::is_view_ { false }

private

◆ name_

template<TensorDataType TDataType, typename TMemoryResource>

std::string Mila::Dnn::Tensor< TDataType, TMemoryResource >::name_

private

Optional user-assigned name for debugging.

◆ shape_

template<TensorDataType TDataType, typename TMemoryResource>

shape_t Mila::Dnn::Tensor< TDataType, TMemoryResource >::shape_ {}

private

Dimensional sizes for each tensor dimension.

◆ size_

template<TensorDataType TDataType, typename TMemoryResource>

size_t Mila::Dnn::Tensor< TDataType, TMemoryResource >::size_ { 0 }

private

Total number of logical elements in the tensor.

◆ strides_

template<TensorDataType TDataType, typename TMemoryResource>

stride_t Mila::Dnn::Tensor< TDataType, TMemoryResource >::strides_ {}

private

Memory stride values for multi-dimensional indexing.

◆ uid_

template<TensorDataType TDataType, typename TMemoryResource>

std::string Mila::Dnn::Tensor< TDataType, TMemoryResource >::uid_

private

Unique identifier for this tensor instance.

◆ view_offset_

template<TensorDataType TDataType, typename TMemoryResource>

size_t Mila::Dnn::Tensor< TDataType, TMemoryResource >::view_offset_ { 0 }

private

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

/__w/Mila/Mila/Mila/Src/Dnn/Tensors/Tensor.ixx

Public Types

Public Member Functions

Static Public Member Functions

Private Member Functions

Static Private Member Functions

Private Attributes

Detailed Description

Member Typedef Documentation

◆ DataType

◆ DataTypeTraits

◆ host_value_t

◆ MemoryResource

Constructor & Destructor Documentation

◆ Tensor() [1/4]

◆ Tensor() [2/4]

◆ Tensor() [3/4]

◆ ~Tensor()

◆ Tensor() [4/4]

Member Function Documentation

◆ allocateBuffer()

◆ computeFlatIndex()

◆ computeSize()

◆ computeStrides()

◆ data() [1/2]

◆ data() [2/2]

◆ elementSize()

◆ empty()

◆ flatten()

◆ getBufferString()

◆ getDataType()

◆ getDataTypeName()

◆ getDeviceId()

◆ getDeviceType()

◆ getMemoryResource()

◆ getName()

◆ getStorageSize()

◆ getUId()

◆ is_device_accessible()

◆ is_host_accessible()

◆ isScalar()

◆ isValid()

◆ isView()

◆ item() [1/2]

◆ item() [2/2]

◆ operator=() [1/2]

◆ operator=() [2/2]

◆ operator[]() [1/4]

◆ operator[]() [2/4]

◆ operator[]() [3/4]

◆ operator[]() [4/4]

◆ outputBuffer()

◆ outputLayout()

◆ rank()

◆ rawData() [1/2]

◆ rawData() [2/2]

◆ reshape()

◆ setName()

◆ setUId()

◆ shape()

◆ size()

◆ strides()

◆ toString()

◆ validateDeviceId()

◆ validateIndices()

◆ view()

Member Data Documentation

◆ buffer_

◆ device_id_

◆ is_view_

◆ name_

◆ shape_

◆ size_

◆ strides_

◆ uid_

◆ view_offset_