at::TensorBase Class Reference

#include <TensorBase.h>

Inheritance diagram for at::TensorBase:

Classes
struct	unsafe_borrow_t

Public Types
template<typename T >
using	hook_return_void_t = std::enable_if_t< std::is_void< typename c10::invoke_result_t< T &, TensorBase > >::value, unsigned >
template<typename T >
using	hook_return_var_t = std::enable_if_t< std::is_same< typename c10::invoke_result_t< T &, TensorBase >, TensorBase >::value, unsigned >

Public Member Functions
	TensorBase ()=default
	TensorBase (c10::intrusive_ptr< TensorImpl, UndefinedTensorImpl > tensor_impl)
	TensorBase (const TensorBase &)=default
	TensorBase (TensorBase &&)=default
int64_t	dim () const
int64_t	storage_offset () const
TensorBase	contiguous (MemoryFormat memory_format=MemoryFormat::Contiguous) const
c10::MaybeOwned< TensorBase >	expect_contiguous (MemoryFormat memory_format=MemoryFormat::Contiguous) const &
	Should be used if *this can reasonably be expected to be contiguous and performance is important. More...
c10::MaybeOwned< TensorBase >	expect_contiguous (MemoryFormat memory_format=MemoryFormat::Contiguous) &&=delete
const TensorBase &	fill_ (const c10::Scalar &scalar) const
const TensorBase &	zero_ () const
TensorBase	to (at::TensorOptions options={}, bool non_blocking=false, bool copy=false, c10::optional< at::MemoryFormat > memory_format=c10::nullopt) const
bool	is_complex () const
bool	is_floating_point () const
bool	is_signed () const
c10::SymInt	sym_size (int64_t dim) const
c10::SymInt	sym_stride (int64_t dim) const
int64_t	size (int64_t dim) const
int64_t	stride (int64_t dim) const
TensorImpl *	unsafeGetTensorImpl () const
TensorImpl *	unsafeReleaseTensorImpl ()
const c10::intrusive_ptr< TensorImpl, UndefinedTensorImpl > &	getIntrusivePtr () const
c10::intrusive_ptr< TensorImpl, UndefinedTensorImpl >	unsafeReleaseIntrusivePtr ()
bool	defined () const
void	reset ()
TensorBase &	operator= (const TensorBase &x) &
TensorBase &	operator= (TensorBase &&x) &noexcept
TensorBase &	operator= (const TensorBase &) &&=delete
TensorBase &	operator= (TensorBase &&) &&noexcept=delete
bool	is_same (const TensorBase &other) const noexcept
size_t	use_count () const noexcept
size_t	weak_use_count () const noexcept
std::string	toString () const
IntArrayRef	sizes () const
c10::SymIntArrayRef	sym_sizes () const
c10::SymIntArrayRef	sym_strides () const
IntArrayRef	strides () const
c10::optional< DimnameList >	opt_names () const
DimnameList	names () const
int64_t	ndimension () const
bool	is_contiguous (at::MemoryFormat memory_format=at::MemoryFormat::Contiguous) const
bool	is_non_overlapping_and_dense () const
at::MemoryFormat	suggest_memory_format (bool channels_last_strides_exact_match=false) const
size_t	nbytes () const
c10::SymInt	sym_nbytes () const
int64_t	numel () const
c10::SymInt	sym_numel () const
c10::SymInt	sym_storage_offset () const
size_t	itemsize () const
int64_t	element_size () const
DispatchKeySet	key_set () const
ScalarType	scalar_type () const
bool	has_storage () const
const Storage &	storage () const
bool	is_alias_of (const at::TensorBase &other) const
bool	_is_zerotensor () const
void	_set_zero (bool zero) const
bool	is_conj () const
void	_set_conj (bool conjugate) const
bool	is_neg () const
void	_set_neg (bool negative) const
Layout	layout () const
	Returns a Tensor's layout. More...
caffe2::TypeMeta	dtype () const
	Returns a Tensor's dtype (TypeMeta). More...
Device	device () const
	Returns a Tensor's device. More...
int64_t	get_device () const
	Returns a Tensor's device index. More...
bool	is_cpu () const
	Returns if a Tensor has CPU backend. More...
bool	is_cuda () const
	Returns if a Tensor has CUDA backend. More...
bool	is_ipu () const
	Returns if a Tensor has IPU backend. More...
bool	is_xpu () const
	Returns if a Tensor has XPU backend. More...
bool	is_xla () const
	Returns if a Tensor has XLA backend. More...
bool	is_hpu () const
	Returns if a Tensor has HPU backend. More...
bool	is_lazy () const
	Returns if a Tensor has Lazy backend. More...
bool	is_hip () const
	Returns if a Tensor has HIP backend. More...
bool	is_ve () const
	Returns if a Tensor has VE backend. More...
bool	is_sparse () const
	Returns if a Tensor has sparse backend. More...
bool	is_sparse_csr () const
	Returns is a Tensor has a sparse CSR backend. More...
bool	is_mkldnn () const
	Returns if a Tensor is mkldnn tensor. More...
bool	is_mps () const
	Returns if a Tensor is mps tensor. More...
bool	is_ort () const
	Returns if a Tensor is ort tensor. More...
bool	is_vulkan () const
	Returns if a Tensor is vulkan tensor. More...
bool	is_metal () const
	Returns if a Tensor is metal tensor. More...
bool	is_quantized () const
	Returns if a Tensor has quantized backend. More...
bool	is_meta () const
	Returns if a Tensor is a meta tensor. More...
bool	is_inference () const
	Returns if a Tensor is an inference tensor. More...
bool	is_nested () const
QuantizerPtr	quantizer () const
	If a tensor is a quantized tensor, returns its quantizer TODO: it's not in native_functions.yaml yet as it's not exposed to python. More...
bool	has_names () const
	Returns if a Tensor has any dimension names. More...
const NamedTensorMeta *	get_named_tensor_meta () const
	Returns a Tensor's dimension names data structure. More...
NamedTensorMeta *	get_named_tensor_meta ()
TensorOptions	options () const
	Returns the TensorOptions corresponding to this Tensor. More...
void *	data_ptr () const
template<typename T >
T *	data_ptr () const
void	print () const
template<typename T , size_t N>
TensorAccessor< T, N >	accessor () const &
template<typename T , size_t N>
TensorAccessor< T, N >	accessor () &&=delete
template<typename T , size_t N, template< typename U > class PtrTraits = DefaultPtrTraits, typename index_t = int64_t>
GenericPackedTensorAccessor< T, N, PtrTraits, index_t >	generic_packed_accessor () const &
template<typename T , size_t N, template< typename U > class PtrTraits = DefaultPtrTraits, typename index_t = int64_t>
GenericPackedTensorAccessor< T, N >	generic_packed_accessor () &&=delete
template<typename T , size_t N, template< typename U > class PtrTraits = DefaultPtrTraits>
PackedTensorAccessor32< T, N, PtrTraits >	packed_accessor32 () const &
template<typename T , size_t N, template< typename U > class PtrTraits = DefaultPtrTraits>
PackedTensorAccessor32< T, N, PtrTraits >	packed_accessor32 () &&=delete
template<typename T , size_t N, template< typename U > class PtrTraits = DefaultPtrTraits>
PackedTensorAccessor64< T, N, PtrTraits >	packed_accessor64 () const &
template<typename T , size_t N, template< typename U > class PtrTraits = DefaultPtrTraits>
PackedTensorAccessor64< T, N, PtrTraits >	packed_accessor64 () &&=delete
const TensorBase &	set_requires_grad (bool requires_grad) const
bool	requires_grad () const
const Tensor &	_fw_grad (uint64_t level) const
	This function returns the forward gradient for this Tensor at the given level. More...
void	_set_fw_grad (const TensorBase &new_grad, uint64_t level, bool is_inplace_op) const
	This function can be used to set the value of the forward grad. More...
at::TensorBase	tensor_data () const
	NOTE: This is similar to the legacy .data() function on Variable, and is intended to be used from functions that need to access the Variable's equivalent Tensor (i.e. More...
at::TensorBase	variable_data () const
	NOTE: var.variable_data() in C++ has the same semantics as tensor.data in Python, which create a new Variable that shares the same storage and tensor metadata with the original Variable, but with a completely new autograd history. More...
const std::shared_ptr< torch::autograd::Node > &	grad_fn () const
	Gets the gradient function of the Variable. More...
template<typename T >
hook_return_void_t< T >	register_hook (T &&hook) const
	Registers a backward hook. More...
template<typename T >
hook_return_var_t< T >	register_hook (T &&hook) const
void	remove_hook (unsigned pos) const
	Remove hook at given position. More...
bool	is_leaf () const
	All Tensors that have requires_grad() which is false will be leaf Tensors by convention. More...
int64_t	output_nr () const
void	set_data (const TensorBase &new_data) const
TensorBase	data () const
int64_t	_version () const
void	retain_grad () const
	Enables this Tensor to have their :attr:grad populated during :func:backward. More...
bool	retains_grad () const
	Is true if this Tensor is non-leaf and its :attr:grad is enabled to be populated during :func:backward, false otherwise. More...
const TensorBase &	requires_grad_ (bool _requires_grad=true) const
bool	is_view () const
	Returns true if this Variable is a view of another Variable. More...
const TensorBase &	_base () const
	Returns the Variable that this Variable is a view of. More...
const std::string &	name () const
template<typename T >
auto	register_hook (T &&hook) const -> TensorBase::hook_return_void_t< T >

Static Public Member Functions
static TensorBase	wrap_tensor_impl (c10::intrusive_ptr< TensorImpl, UndefinedTensorImpl > tensor_impl)

Protected Member Functions
	TensorBase (unsafe_borrow_t, const TensorBase &rhs)
unsigned	_register_hook (std::function< TensorBase(const TensorBase &)> hook) const
void	enforce_invariants ()

Protected Attributes
friend	MaybeOwnedTraits< TensorBase >
c10::intrusive_ptr< TensorImpl, UndefinedTensorImpl >	impl_

Member Typedef Documentation

hook_return_var_t

template<typename T >

using at::TensorBase::hook_return_var_t = std::enable_if_t<std::is_same<typename c10::invoke_result_t<T&, TensorBase>, TensorBase>::value, unsigned>

hook_return_void_t

template<typename T >

using at::TensorBase::hook_return_void_t = std::enable_if_t<std::is_void<typename c10::invoke_result_t<T&, TensorBase> >::value, unsigned>

Constructor & Destructor Documentation

TensorBase() [1/5]

at::TensorBase::TensorBase ( unsafe_borrow_t  , const TensorBase &  rhs  )

inlineexplicitprotected

TensorBase() [2/5]

at::TensorBase::TensorBase ( )

default

TensorBase() [3/5]

at::TensorBase::TensorBase ( c10::intrusive_ptr< TensorImpl, UndefinedTensorImpl >  tensor_impl )

inlineexplicit

TensorBase() [4/5]

at::TensorBase::TensorBase ( const TensorBase &  )

default

TensorBase() [5/5]

at::TensorBase::TensorBase ( TensorBase &&  )

default

Member Function Documentation

_base()

const TensorBase & at::TensorBase::_base

(

)

const

Returns the Variable that this Variable is a view of.

If this Variable is not a view, throw a std::runtime_error.

_fw_grad()

const Tensor & at::TensorBase::_fw_grad ( uint64_t  level ) const

inline

This function returns the forward gradient for this Tensor at the given level.

_is_zerotensor()

bool at::TensorBase::_is_zerotensor ( ) const

inline

_register_hook()

unsigned at::TensorBase::_register_hook ( std::function< TensorBase(const TensorBase &)>  hook ) const

protected

_set_conj()

void at::TensorBase::_set_conj ( bool  conjugate ) const

inline

_set_fw_grad()

void at::TensorBase::_set_fw_grad ( const TensorBase &  new_grad, uint64_t  level, bool  is_inplace_op  ) const

inline

This function can be used to set the value of the forward grad.

Note that the given new_grad might not be used directly if it has different metadata (size/stride/storage offset) compared to this Tensor. In that case, new_grad content will be copied into a new Tensor

_set_neg()

void at::TensorBase::_set_neg ( bool  negative ) const

inline

_set_zero()

void at::TensorBase::_set_zero ( bool  zero ) const

inline

_version()

int64_t at::TensorBase::_version

(

)

const

accessor() [1/2]

template<typename T , size_t N>

TensorAccessor< T, N > at::TensorBase::accessor ( ) &&

delete

accessor() [2/2]

template<typename T , size_t N>

TensorAccessor< T, N > at::TensorBase::accessor ( ) const &

inline

contiguous()

TensorBase at::TensorBase::contiguous ( MemoryFormat  memory_format = MemoryFormat::Contiguous ) const

inline

data()

TensorBase at::TensorBase::data

(

)

const

data_ptr() [1/2]

void * at::TensorBase::data_ptr ( ) const

inline

data_ptr() [2/2]

template<typename T >

T * at::TensorBase::data_ptr

(

)

const

defined()

bool at::TensorBase::defined ( ) const

inline

device()

Device at::TensorBase::device ( ) const

inline

Returns a Tensor's device.

dim()

int64_t at::TensorBase::dim ( ) const

inline

dtype()

caffe2::TypeMeta at::TensorBase::dtype ( ) const

inline

Returns a Tensor's dtype (TypeMeta).

element_size()

int64_t at::TensorBase::element_size ( ) const

inline

enforce_invariants()

void at::TensorBase::enforce_invariants ( )

protected

expect_contiguous() [1/2]

c10::MaybeOwned< TensorBase > at::TensorBase::expect_contiguous ( MemoryFormat  memory_format = MemoryFormat::Contiguous ) &&

delete

expect_contiguous() [2/2]

c10::MaybeOwned< TensorBase > at::TensorBase::expect_contiguous ( MemoryFormat  memory_format = MemoryFormat::Contiguous ) const &

inline

Should be used if *this can reasonably be expected to be contiguous and performance is important.

Compared to contiguous, it saves a reference count increment/decrement if *this is already contiguous, at the cost in all cases of an extra pointer of stack usage, an extra branch to access, and an extra branch at destruction time.

fill_()

const TensorBase & at::TensorBase::fill_

(

const c10::Scalar &

scalar

)

const

generic_packed_accessor() [1/2]

template<typename T , size_t N, template< typename U > class PtrTraits = DefaultPtrTraits, typename index_t = int64_t>

GenericPackedTensorAccessor< T, N > at::TensorBase::generic_packed_accessor ( ) &&

delete

generic_packed_accessor() [2/2]

template<typename T , size_t N, template< typename U > class PtrTraits = DefaultPtrTraits, typename index_t = int64_t>

GenericPackedTensorAccessor< T, N, PtrTraits, index_t > at::TensorBase::generic_packed_accessor ( ) const &

inline

get_device()

int64_t at::TensorBase::get_device ( ) const

inline

Returns a Tensor's device index.

get_named_tensor_meta() [1/2]

NamedTensorMeta * at::TensorBase::get_named_tensor_meta ( )

inline

get_named_tensor_meta() [2/2]

const NamedTensorMeta * at::TensorBase::get_named_tensor_meta ( ) const

inline

Returns a Tensor's dimension names data structure.

getIntrusivePtr()

const c10::intrusive_ptr< TensorImpl, UndefinedTensorImpl > & at::TensorBase::getIntrusivePtr ( ) const

inline

grad_fn()

const std::shared_ptr< torch::autograd::Node > & at::TensorBase::grad_fn

(

)

const

Gets the gradient function of the Variable.

If this is a leaf variable, the pointer returned will be null.

For View Variables: Gets the up-to-date grad_fn. If the shared data or base was modified, we re-create the grad_fn to express the up-to-date view relationship between this and the base Variable.

has_names()

bool at::TensorBase::has_names ( ) const

inline

Returns if a Tensor has any dimension names.

has_storage()

bool at::TensorBase::has_storage ( ) const

inline

is_alias_of()

bool at::TensorBase::is_alias_of ( const at::TensorBase &  other ) const

inline

is_complex()

bool at::TensorBase::is_complex ( ) const

inline

is_conj()

bool at::TensorBase::is_conj ( ) const

inline

is_contiguous()

bool at::TensorBase::is_contiguous ( at::MemoryFormat  memory_format = at::MemoryFormat::Contiguous ) const

inline

is_cpu()

bool at::TensorBase::is_cpu ( ) const

inline

Returns if a Tensor has CPU backend.

is_cuda()

bool at::TensorBase::is_cuda ( ) const

inline

Returns if a Tensor has CUDA backend.

is_floating_point()

bool at::TensorBase::is_floating_point ( ) const

inline

is_hip()

bool at::TensorBase::is_hip ( ) const

inline

Returns if a Tensor has HIP backend.

is_hpu()

bool at::TensorBase::is_hpu ( ) const

inline

Returns if a Tensor has HPU backend.

is_inference()

bool at::TensorBase::is_inference ( ) const

inline

Returns if a Tensor is an inference tensor.

is_ipu()

bool at::TensorBase::is_ipu ( ) const

inline

Returns if a Tensor has IPU backend.

is_lazy()

bool at::TensorBase::is_lazy ( ) const

inline

Returns if a Tensor has Lazy backend.

is_leaf()

bool at::TensorBase::is_leaf

(

)

const

All Tensors that have requires_grad() which is false will be leaf Tensors by convention.

For Tensors that have requires_grad() which is true, they will be leaf Tensors if they were created by the user. This means that they are not the result of an operation and so grad_fn() is nullptr.

Only leaf Tensors will have their grad() populated during a call to backward(). To get grad() populated for non-leaf Tensors, you can use retain_grad().

Example:

auto a = torch::rand(10, torch::requires_grad());
std::cout << a.is_leaf() << std::endl; // prints `true`
 
auto b = torch::rand(10, torch::requires_grad()).to(torch::kCUDA);
std::cout << b.is_leaf() << std::endl; // prints `false`
// b was created by the operation that cast a cpu Tensor into a cuda Tensor
 
auto c = torch::rand(10, torch::requires_grad()) + 2;
std::cout << c.is_leaf() << std::endl; // prints `false`
// c was created by the addition operation
 
auto d = torch::rand(10).cuda();
std::cout << d.is_leaf() << std::endl; // prints `true`
// d does not require gradients and so has no operation creating it (that is tracked by the autograd engine)
 
auto e = torch::rand(10).cuda().requires_grad_();
std::cout << e.is_leaf() << std::endl; // prints `true`
// e requires gradients and has no operations creating it
 
auto f = torch::rand(10, torch::device(torch::kCUDA).requires_grad(true));
std::cout << f.is_leaf() << std::endl; // prints `true`
// f requires grad, has no operation creating it

is_meta()

bool at::TensorBase::is_meta ( ) const

inline

Returns if a Tensor is a meta tensor.

Meta tensors can also have other designations.

is_metal()

bool at::TensorBase::is_metal ( ) const

inline

Returns if a Tensor is metal tensor.

is_mkldnn()

bool at::TensorBase::is_mkldnn ( ) const

inline

Returns if a Tensor is mkldnn tensor.

is_mps()

bool at::TensorBase::is_mps ( ) const

inline

Returns if a Tensor is mps tensor.

is_neg()

bool at::TensorBase::is_neg ( ) const

inline

is_nested()

bool at::TensorBase::is_nested ( ) const

inline

is_non_overlapping_and_dense()

bool at::TensorBase::is_non_overlapping_and_dense ( ) const

inline

is_ort()

bool at::TensorBase::is_ort ( ) const

inline

Returns if a Tensor is ort tensor.

is_quantized()

bool at::TensorBase::is_quantized ( ) const

inline

Returns if a Tensor has quantized backend.

is_same()

bool at::TensorBase::is_same ( const TensorBase &  other ) const

inlinenoexcept

is_signed()

bool at::TensorBase::is_signed ( ) const

inline

is_sparse()

bool at::TensorBase::is_sparse ( ) const

inline

Returns if a Tensor has sparse backend.

is_sparse_csr()

bool at::TensorBase::is_sparse_csr ( ) const

inline

Returns is a Tensor has a sparse CSR backend.

is_ve()

bool at::TensorBase::is_ve ( ) const

inline

Returns if a Tensor has VE backend.

is_view()

bool at::TensorBase::is_view

(

)

const

Returns true if this Variable is a view of another Variable.

is_vulkan()

bool at::TensorBase::is_vulkan ( ) const

inline

Returns if a Tensor is vulkan tensor.

is_xla()

bool at::TensorBase::is_xla ( ) const

inline

Returns if a Tensor has XLA backend.

is_xpu()

bool at::TensorBase::is_xpu ( ) const

inline

Returns if a Tensor has XPU backend.

itemsize()

size_t at::TensorBase::itemsize ( ) const

inline

key_set()

DispatchKeySet at::TensorBase::key_set ( ) const

inline

layout()

Layout at::TensorBase::layout ( ) const

inline

Returns a Tensor's layout.

name()

const std::string & at::TensorBase::name

(

)

const

names()

DimnameList at::TensorBase::names ( ) const

inline

nbytes()

size_t at::TensorBase::nbytes ( ) const

inline

ndimension()

int64_t at::TensorBase::ndimension ( ) const

inline

numel()

int64_t at::TensorBase::numel ( ) const

inline

operator=() [1/4]

TensorBase & at::TensorBase::operator= ( const TensorBase &  ) &&

delete

operator=() [2/4]

TensorBase & at::TensorBase::operator= ( const TensorBase &  x ) &

inline

operator=() [3/4]

TensorBase & at::TensorBase::operator= ( TensorBase &&  ) &&

deletenoexcept

operator=() [4/4]

TensorBase & at::TensorBase::operator= ( TensorBase &&  x ) &

inlinenoexcept

opt_names()

c10::optional< DimnameList > at::TensorBase::opt_names ( ) const

inline

options()

TensorOptions at::TensorBase::options ( ) const

inline

Returns the TensorOptions corresponding to this Tensor.

Defined in TensorOptions.h.

output_nr()

int64_t at::TensorBase::output_nr

(

)

const

packed_accessor32() [1/2]

template<typename T , size_t N, template< typename U > class PtrTraits = DefaultPtrTraits>

PackedTensorAccessor32< T, N, PtrTraits > at::TensorBase::packed_accessor32 ( ) &&

delete

packed_accessor32() [2/2]

template<typename T , size_t N, template< typename U > class PtrTraits = DefaultPtrTraits>

PackedTensorAccessor32< T, N, PtrTraits > at::TensorBase::packed_accessor32 ( ) const &

inline

packed_accessor64() [1/2]

template<typename T , size_t N, template< typename U > class PtrTraits = DefaultPtrTraits>

PackedTensorAccessor64< T, N, PtrTraits > at::TensorBase::packed_accessor64 ( ) &&

delete

packed_accessor64() [2/2]

template<typename T , size_t N, template< typename U > class PtrTraits = DefaultPtrTraits>

PackedTensorAccessor64< T, N, PtrTraits > at::TensorBase::packed_accessor64 ( ) const &

inline

print()

void at::TensorBase::print

(

)

const

quantizer()

QuantizerPtr at::TensorBase::quantizer

(

)

const

If a tensor is a quantized tensor, returns its quantizer TODO: it's not in native_functions.yaml yet as it's not exposed to python.

register_hook() [1/3]

template<typename T >

hook_return_void_t< T > at::TensorBase::register_hook

(

T &&

hook

)

const

Registers a backward hook.

The hook will be called every time a gradient with respect to the Tensor is computed. The hook should have one of the following signature:

hook(TensorBase grad) -> TensorBase

hook(TensorBase grad) -> void

The hook should not modify its argument, but it can optionally return a new gradient which will be used in place of grad.

This function returns the index of the hook in the list which can be used to remove hook.

Example:

auto v = torch::tensor({0., 0., 0.}, torch::requires_grad());
auto h = v.register_hook([](torch::Tensor grad){ return grad * 2; }); // double the gradient
v.backward(torch::tensor({1., 2., 3.}));
// This prints:
// ```
//  2
//  4
//  6
// [ CPUFloatType{3} ]
// ```
std::cout << v.grad() << std::endl;
v.remove_hook(h);  // removes the hook

register_hook() [2/3]

template<typename T >

hook_return_var_t< T > at::TensorBase::register_hook

(

T &&

hook

)

const

register_hook() [3/3]

template<typename T >

auto at::TensorBase::register_hook

(

T &&

hook

)

const -> TensorBase::hook_return_void_t<T>

remove_hook()

void at::TensorBase::remove_hook

(

unsigned

pos

)

const

Remove hook at given position.

requires_grad()

bool at::TensorBase::requires_grad ( ) const

inline

requires_grad_()

const TensorBase & at::TensorBase::requires_grad_

(

bool

_requires_grad = true

)

const

reset()

void at::TensorBase::reset ( )

inline

retain_grad()

void at::TensorBase::retain_grad

(

)

const

Enables this Tensor to have their :attr:grad populated during :func:backward.

This is a no-op for leaf tensors.

retains_grad()

bool at::TensorBase::retains_grad

(

)

const

Is true if this Tensor is non-leaf and its :attr:grad is enabled to be populated during :func:backward, false otherwise.

scalar_type()

ScalarType at::TensorBase::scalar_type ( ) const

inline

set_data()

void at::TensorBase::set_data

(

const TensorBase &

new_data

)

const

set_requires_grad()

const TensorBase & at::TensorBase::set_requires_grad ( bool  requires_grad ) const

inline

size()

int64_t at::TensorBase::size ( int64_t  dim ) const

inline

sizes()

IntArrayRef at::TensorBase::sizes ( ) const

inline

storage()

const Storage & at::TensorBase::storage ( ) const

inline

storage_offset()

int64_t at::TensorBase::storage_offset ( ) const

inline

stride()

int64_t at::TensorBase::stride ( int64_t  dim ) const

inline

strides()

IntArrayRef at::TensorBase::strides ( ) const

inline

suggest_memory_format()

at::MemoryFormat at::TensorBase::suggest_memory_format ( bool  channels_last_strides_exact_match = false ) const

inline

sym_nbytes()

c10::SymInt at::TensorBase::sym_nbytes ( ) const

inline

sym_numel()

c10::SymInt at::TensorBase::sym_numel ( ) const

inline

sym_size()

c10::SymInt at::TensorBase::sym_size ( int64_t  dim ) const

inline

sym_sizes()

c10::SymIntArrayRef at::TensorBase::sym_sizes ( ) const

inline

sym_storage_offset()

c10::SymInt at::TensorBase::sym_storage_offset ( ) const

inline

sym_stride()

c10::SymInt at::TensorBase::sym_stride ( int64_t  dim ) const

inline

sym_strides()

c10::SymIntArrayRef at::TensorBase::sym_strides ( ) const

inline

tensor_data()

at::TensorBase at::TensorBase::tensor_data

(

)

const

NOTE: This is similar to the legacy .data() function on Variable, and is intended to be used from functions that need to access the Variable's equivalent Tensor (i.e.

Tensor that shares the same storage and tensor metadata with the Variable).

One notable difference with the legacy .data() function is that changes to the returned Tensor's tensor metadata (e.g. sizes / strides / storage / storage_offset) will not update the original Variable, due to the fact that this function shallow-copies the Variable's underlying TensorImpl.

to()

TensorBase at::TensorBase::to	(	at::TensorOptions	options = {},
		bool	non_blocking = false,
		bool	copy = false,
		c10::optional< at::MemoryFormat >	memory_format = c10::nullopt
	)		const

toString()

std::string at::TensorBase::toString

(

)

const

unsafeGetTensorImpl()

TensorImpl * at::TensorBase::unsafeGetTensorImpl ( ) const

inline

unsafeReleaseIntrusivePtr()

c10::intrusive_ptr< TensorImpl, UndefinedTensorImpl > at::TensorBase::unsafeReleaseIntrusivePtr ( )

inline

unsafeReleaseTensorImpl()

TensorImpl * at::TensorBase::unsafeReleaseTensorImpl ( )

inline

use_count()

size_t at::TensorBase::use_count ( ) const

inlinenoexcept

variable_data()

at::TensorBase at::TensorBase::variable_data

(

)

const

NOTE: var.variable_data() in C++ has the same semantics as tensor.data in Python, which create a new Variable that shares the same storage and tensor metadata with the original Variable, but with a completely new autograd history.

NOTE: If we change the tensor metadata (e.g. sizes / strides / storage / storage_offset) of a variable created from var.variable_data(), those changes will not update the original variable var. In .variable_data(), we set allow_tensor_metadata_change_ to false to make such changes explicitly illegal, in order to prevent users from changing metadata of var.variable_data() and expecting the original variable var to also be updated.

weak_use_count()

size_t at::TensorBase::weak_use_count ( ) const

inlinenoexcept

wrap_tensor_impl()

static TensorBase at::TensorBase::wrap_tensor_impl ( c10::intrusive_ptr< TensorImpl, UndefinedTensorImpl >  tensor_impl )

inlinestatic

zero_()

const TensorBase & at::TensorBase::zero_

(

)

const

Member Data Documentation

impl_

c10::intrusive_ptr<TensorImpl, UndefinedTensorImpl> at::TensorBase::impl_

protected

MaybeOwnedTraits< TensorBase >

friend at::TensorBase::MaybeOwnedTraits< TensorBase >

protected

---

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

• aten/src/ATen/core/TensorBase.h