cldnn::tensor Struct Reference

N-dimensional vector. Mostly used to represent memory size. More...

#include <tensor.hpp>

Public Types
typedef int32_t	value_type
	Values type stored in tensor.

Public Member Functions
	tensor (value_type default_size=0)
template<typename ... KindInitTys, typename = typename std::enable_if< meta::all< meta::is_any_of<KindInitTys, cldnn::details::dim_vec_kind_init<cldnn::details::dim_vec_kind::batch>, cldnn::details::dim_vec_kind_init<cldnn::details::dim_vec_kind::feature>, cldnn::details::dim_vec_kind_init<details::dim_vec_kind::spatial> >::value... >::value, void>::type>
	tensor (KindInitTys &&... kind_inits)
	Constructs tensor. More...
	tensor (value_type batch_num, value_type feature_num, value_type width, value_type height)
	Constructs tensor. More...
	tensor (const std::vector< value_type > &sizes, value_type default_size=1)
	Constructs tensor using vector of sizes. More...
	tensor (format fmt, const std::vector< value_type > &sizes, value_type default_size=1)
	tensor (const cldnn_tensor &other)
	Implicit conversion form C API :: cldnn_tensor.
	operator cldnn_tensor () const
	Implicit conversion to C API cldnn_tensor.
	tensor (const tensor &other)
	Copy construction.
tensor &	operator= (const tensor &other)
	Copy assignment.
std::string	to_string () const
tensor	negate () const
	Returns a tensor with all negated elements.
tensor	mul (value_type multiplier) const
	Returns a tensor with all elements multilied to multiplier.
tensor	div (value_type divider) const
	Returns a tensor with all elements divided by divider.
tensor	add (const tensor &rhs) const
	Returns a tensor with all elements added by appropriate elements of rhs.
tensor	sub (const tensor &rhs) const
	Returns a tensor with all elements subtracted by appropriate elements of rhs.
tensor &	operator+= (const tensor &rhs)
	Assign and add.
tensor &	operator-= (const tensor &rhs)
	Assign and subtract.
std::vector< value_type >	sizes (cldnn::format fmt) const
	Returns a vector of tensors values, ordered regarding to format.
std::vector< value_type >	sizes () const
	Returns a vector of tensors values, ordered batch, feature, spatial_x, spatial_y.
size_t	count () const
	Returns tensor elements count calculated as multiplication of all elements.
tensor	transform (cldnn::format new_fmt, value_type default_size) const
	Returns new tensor based on current but transformed to new format. More...
size_t	get_linear_offset (const tensor &coord, cldnn::format fmt) const
	Calculates linear offset for given coord within current tensor. More...

Static Public Member Functions
static tensor	max (tensor const &lhs, tensor const &rhs)
	Returns a tensor containing values maximum from lhs and rhs.
static tensor	min (tensor const &lhs, tensor const &rhs)
	Returns a tensor containing values minimum from lhs and rhs.

Public Attributes
mutable_array_ref< value_type >	raw
	Raw representation of all dimensions.
mutable_array_ref< value_type >	batch
	Batch dimensions.
mutable_array_ref< value_type >	feature
	Feature maps.
mutable_array_ref< value_type >	spatial
	Spatial dimensions.

Friends
bool	operator== (const tensor &lhs, const tensor &rhs)
bool	operator!= (const tensor &lhs, const tensor &rhs)
bool	operator< (const tensor &lhs, const tensor &rhs)
std::ostream &	operator<< (std::ostream &os, const tensor &tensor)

Detailed Description

N-dimensional vector. Mostly used to represent memory size.

Definition at line 256 of file tensor.hpp.

Constructor & Destructor Documentation

tensor() [1/3]

template<typename ... KindInitTys, typename = typename std::enable_if< meta::all< meta::is_any_of<KindInitTys, cldnn::details::dim_vec_kind_init<cldnn::details::dim_vec_kind::batch>, cldnn::details::dim_vec_kind_init<cldnn::details::dim_vec_kind::feature>, cldnn::details::dim_vec_kind_init<details::dim_vec_kind::spatial> >::value... >::value, void>::type>

cldnn::tensor::tensor ( KindInitTys &&...  kind_inits )

inline

Constructs tensor.

Parameters

[in]

kind_inits

Dimensions defined using dim_vec_kind. If dimension is not provided it is set to 1.

Example:

tensor my_tensor(batch(2), spatial(5, 6));   // y=6, x=5, b=2, f - not set
cout << my_tensor.batch[0] << endl;           // 2
cout << my_tensor.feature[0] << endl;         // 1 - default_size
cout << "x=" << my_tensor.spatial[0] << endl; // x=5
cout << "y=" << my_tensor.spatial[1] << endl; // y=6

Definition at line 306 of file tensor.hpp.

tensor() [2/3]

cldnn::tensor::tensor ( value_type  batch_num, value_type  feature_num, value_type  width, value_type  height  )

inline

Constructs tensor.

Example:

tensor my_tensor( 2, 3, 4, 5 );   // b=2, f=3, x=4, y=5
cout << my_tensor.batch[0] << endl;           // 2
cout << my_tensor.feature[0] << endl;         // 3
cout << "x=" << my_tensor.spatial[0] << endl; // x=4
cout << "y=" << my_tensor.spatial[1] << endl; // y=5

Definition at line 324 of file tensor.hpp.

tensor() [3/3]

cldnn::tensor::tensor ( const std::vector< value_type > &  sizes, value_type  default_size = 1  )

inline

Constructs tensor using vector of sizes.

Parameters

[in]	sizes	dimensions need to be provided in the following order {batch, feature, spatial_x, spatial_y}.
[in]	default_size	default_size for tensor dimensions.

Example:

tensor my_tensor = { 2, 3, 4, 5 };   // b=2, f=3, x=4, y=5
cout << my_tensor.batch[0] << endl;           // 2
cout << my_tensor.feature[0] << endl;         // 3
cout << "x=" << my_tensor.spatial[0] << endl; // x=4
cout << "y=" << my_tensor.spatial[1] << endl; // y=5

Definition at line 347 of file tensor.hpp.

Member Function Documentation

get_linear_offset()

size_t cldnn::tensor::get_linear_offset ( const tensor &  coord, cldnn::format  fmt  ) const

inline

Calculates linear offset for given coord within current tensor.

Parameters

coord

The coordinate within current tensor.

Definition at line 639 of file tensor.hpp.

transform()

tensor cldnn::tensor::transform ( cldnn::format  new_fmt, value_type  default_size  ) const

inline

Returns new tensor based on current but transformed to new format.

Parameters

[in]	new_fmt	Format of new tensor.
[in]	default_size	Default element values for positions not defined by current format.

Example:

tensor my_tensor({ 2, 3, 4, 5 });
auto my_sizes = my_tensor.sizes();
cout << "dims_num=" << my_sizes.size() << endl; // dims_num=2
cout << "b=" << my_sizes[0] << endl;            // b=2
cout << "f=" << my_sizes[1] << endl;            // f=3
cout << "x=" << my_sizes[2] << endl;            // x=5
cout << "y=" << my_sizes[3] << endl;            // y=4
auto new_tensor = my_tensor.transform(format::yxfb, 10);
auto new_sizes = new_tensor.sizes();
cout << "new_num=" << new_sizes.size() << endl;   // new_num=4
for(auto dim : new_sizes) cout << " " << dim;     //  5 4 3 2
cout << endl;

Definition at line 593 of file tensor.hpp.

---

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

• C:/ML/OpenSource/clDNN/api/CPP/tensor.hpp