# torch.Tensor¶

A torch.Tensor is a multi-dimensional matrix containing elements of a single data type.

Torch defines eight CPU tensor types and eight GPU tensor types:

Data type dtype CPU tensor GPU tensor
32-bit floating point torch.float32 or torch.float torch.FloatTensor torch.cuda.FloatTensor
64-bit floating point torch.float64 or torch.double torch.DoubleTensor torch.cuda.DoubleTensor
16-bit floating point torch.float16 or torch.half torch.HalfTensor torch.cuda.HalfTensor
8-bit integer (unsigned) torch.uint8 torch.ByteTensor torch.cuda.ByteTensor
8-bit integer (signed) torch.int8 torch.CharTensor torch.cuda.CharTensor
16-bit integer (signed) torch.int16 or torch.short torch.ShortTensor torch.cuda.ShortTensor
32-bit integer (signed) torch.int32 or torch.int torch.IntTensor torch.cuda.IntTensor
64-bit integer (signed) torch.int64 or torch.long torch.LongTensor torch.cuda.LongTensor

torch.Tensor is an alias for the default tensor type (torch.FloatTensor).

A tensor can be constructed from a Python list or sequence using the torch.tensor() constructor:

>>> torch.tensor([[1., -1.], [1., -1.]])
tensor([[ 1.0000, -1.0000],
[ 1.0000, -1.0000]])
>>> torch.tensor(np.array([[1, 2, 3], [4, 5, 6]]))
tensor([[ 1,  2,  3],
[ 4,  5,  6]])


Warning

torch.tensor() always copies data. If you have a Tensor data and just want to change its requires_grad flag, use requires_grad_() or detach() to avoid a copy. If you have a numpy array and want to avoid a copy, use torch.from_numpy().

An tensor of specific data type can be constructed by passing a torch.dtype and/or a torch.device to a constructor or tensor creation op:

>>> torch.zeros([2, 4], dtype=torch.int32)
tensor([[ 0,  0,  0,  0],
[ 0,  0,  0,  0]], dtype=torch.int32)
>>> cuda0 = torch.device('cuda:0')
>>> torch.ones([2, 4], dtype=torch.float64, device=cuda0)
tensor([[ 1.0000,  1.0000,  1.0000,  1.0000],
[ 1.0000,  1.0000,  1.0000,  1.0000]], dtype=torch.float64, device='cuda:0')


The contents of a tensor can be accessed and modified using Python’s indexing and slicing notation:

>>> x = torch.tensor([[1, 2, 3], [4, 5, 6]])
>>> print(x[1][2])
tensor(6)
>>> x[0][1] = 8
>>> print(x)
tensor([[ 1,  8,  3],
[ 4,  5,  6]])


Use torch.Tensor.item() to get a Python number from a tensor containing a single value:

>>> x = torch.tensor([[1]])
>>> x
tensor([[ 1]])
>>> x.item()
1
>>> x = torch.tensor(2.5)
>>> x
tensor(2.5000)
>>> x.item()
2.5


A tensor can be created with requires_grad=True so that torch.autograd records operations on them for automatic differentiation.

>>> x = torch.tensor([[1., -1.], [1., 1.]], requires_grad=True)
>>> out = x.pow(2).sum()
>>> out.backward()
tensor([[ 2.0000, -2.0000],
[ 2.0000,  2.0000]])


Each tensor has an associated torch.Storage, which holds its data. The tensor class provides multi-dimensional, strided view of a storage and defines numeric operations on it.

Note

Methods which mutate a tensor are marked with an underscore suffix. For example, torch.FloatTensor.abs_() computes the absolute value in-place and returns the modified tensor, while torch.FloatTensor.abs() computes the result in a new tensor.

Note

To change an existing tensor’s torch.device and/or torch.dtype, consider using to() method on the tensor.

class torch.Tensor

There are a few main ways to create a tensor, depending on your use case.

• To create a tensor with pre-existing data, use torch.tensor().
• To create a tensor with specific size, use torch.* tensor creation ops (see Creation Ops).
• To create a tensor with the same size (and similar types) as another tensor, use torch.*_like tensor creation ops (see Creation Ops).
• To create a tensor with similar type but different size as another tensor, use tensor.new_* creation ops.
new_tensor(data, dtype=None, device=None, requires_grad=False) → Tensor

Returns a new Tensor with data as the tensor data. By default, the returned Tensor has the same torch.dtype and torch.device as this tensor.

Warning

new_tensor() always copies data. If you have a Tensor data and want to avoid a copy, use torch.Tensor.requires_grad_() or torch.Tensor.detach(). If you have a numpy array and want to avoid a copy, use torch.from_numpy().

Parameters: data (array_like) – The returned Tensor copies data. dtype (torch.dtype, optional) – the desired type of returned tensor. device (torch.device, optional) – the desired device of returned tensor. requires_grad (bool, optional) – If autograd should record operations on the

Example:

>>> tensor = torch.ones((2,), dtype=torch.int8)
>>> data = [[0, 1], [2, 3]]
>>> tensor.new_tensor(data)
tensor([[ 0,  1],
[ 2,  3]], dtype=torch.int8)

new_full(size, fill_value, dtype=None, device=None, requires_grad=False) → Tensor

Returns a Tensor of size size filled with fill_value. By default, the returned Tensor has the same torch.dtype and torch.device as this tensor.

Parameters: fill_value (scalar) – the number to fill the output tensor with. dtype (torch.dtype, optional) – the desired type of returned tensor. device (torch.device, optional) – the desired device of returned tensor. requires_grad (bool, optional) – If autograd should record operations on the

Example:

>>> tensor = torch.ones((2,), dtype=torch.float64)
>>> tensor.new_full((3, 4), 3.141592)
tensor([[ 3.1416,  3.1416,  3.1416,  3.1416],
[ 3.1416,  3.1416,  3.1416,  3.1416],
[ 3.1416,  3.1416,  3.1416,  3.1416]], dtype=torch.float64)

new_empty(size, dtype=None, device=None, requires_grad=False) → Tensor

Returns a Tensor of size size filled with uninitialized data. By default, the returned Tensor has the same torch.dtype and torch.device as this tensor.

Parameters: dtype (torch.dtype, optional) – the desired type of returned tensor. device (torch.device, optional) – the desired device of returned tensor. requires_grad (bool, optional) – If autograd should record operations on the

Example:

>>> tensor = torch.ones(())
>>> tensor.new_empty((2, 3))
tensor([[ 5.8182e-18,  4.5765e-41, -1.0545e+30],
[ 3.0949e-41,  4.4842e-44,  0.0000e+00]])

new_ones(size, dtype=None, device=None, requires_grad=False) → Tensor

Returns a Tensor of size size filled with 1. By default, the returned Tensor has the same torch.dtype and torch.device as this tensor.

Parameters: size (int...) – a list, tuple, or torch.Size of integers defining the shape of the output tensor. dtype (torch.dtype, optional) – the desired type of returned tensor. device (torch.device, optional) – the desired device of returned tensor. requires_grad (bool, optional) – If autograd should record operations on the

Example:

>>> tensor = torch.tensor((), dtype=torch.int32)
>>> tensor.new_ones((2, 3))
tensor([[ 1,  1,  1],
[ 1,  1,  1]], dtype=torch.int32)

new_zeros(size, dtype=None, device=None, requires_grad=False) → Tensor

Returns a Tensor of size size filled with 0. By default, the returned Tensor has the same torch.dtype and torch.device as this tensor.

Parameters: size (int...) – a list, tuple, or torch.Size of integers defining the shape of the output tensor. dtype (torch.dtype, optional) – the desired type of returned tensor. device (torch.device, optional) – the desired device of returned tensor. requires_grad (bool, optional) – If autograd should record operations on the

Example:

>>> tensor = torch.tensor((), dtype=torch.float64)
>>> tensor.new_ones((2, 3))
tensor([[ 1.,  1.,  1.],
[ 1.,  1.,  1.]], dtype=torch.float64)

abs() → Tensor
abs_() → Tensor

In-place version of abs()

acos() → Tensor
acos_() → Tensor

In-place version of acos()

add(value) → Tensor
add_(value) → Tensor

In-place version of add()

addbmm(beta=1, mat, alpha=1, batch1, batch2) → Tensor
addbmm_(beta=1, mat, alpha=1, batch1, batch2) → Tensor

In-place version of addbmm()

addcdiv(value=1, tensor1, tensor2) → Tensor
addcdiv_(value=1, tensor1, tensor2) → Tensor

In-place version of addcdiv()

addcmul(value=1, tensor1, tensor2) → Tensor
addcmul_(value=1, tensor1, tensor2) → Tensor

In-place version of addcmul()

addmm(beta=1, mat, alpha=1, mat1, mat2) → Tensor
addmm_(beta=1, mat, alpha=1, mat1, mat2) → Tensor

In-place version of addmm()

addmv(beta=1, tensor, alpha=1, mat, vec) → Tensor
addmv_(beta=1, tensor, alpha=1, mat, vec) → Tensor

In-place version of addmv()

addr(beta=1, alpha=1, vec1, vec2) → Tensor
addr_(beta=1, alpha=1, vec1, vec2) → Tensor

In-place version of addr()

apply_(callable) → Tensor

Applies the function callable to each element in the tensor, replacing each element with the value returned by callable.

Note

This function only works with CPU tensors and should not be used in code sections that require high performance.

argmax(dim=None, keepdim=False)[source]
argmin(dim=None, keepdim=False)[source]
asin() → Tensor
asin_() → Tensor

In-place version of asin()

atan() → Tensor
atan2(other) → Tensor
atan2_(other) → Tensor

In-place version of atan2()

atan_() → Tensor

In-place version of atan()

baddbmm(beta=1, alpha=1, batch1, batch2) → Tensor
baddbmm_(beta=1, alpha=1, batch1, batch2) → Tensor

In-place version of baddbmm()

bernoulli() → Tensor
bernoulli_() → Tensor

In-place version of bernoulli()

bmm(batch2) → Tensor
byte() → Tensor

self.byte() is equivalent to self.to(torch.uint8). See to().

btrifact(info=None, pivot=True)[source]
btrifact_with_info(pivot=True) -> (Tensor, Tensor, Tensor)
btrisolve()
cauchy_(median=0, sigma=1, *, generator=None) → Tensor

Fills the tensor with numbers drawn from the Cauchy distribution:

$f(x) = \dfrac{1}{\pi} \dfrac{\sigma}{(x - median)^2 + \sigma^2}$
ceil() → Tensor
ceil_() → Tensor

In-place version of ceil()

char() → Tensor

self.char() is equivalent to self.to(torch.int8). See to().

chunk(chunks, dim=0) → List of Tensors
clamp(min, max) → Tensor
clamp_(min, max) → Tensor

In-place version of clamp()

clone() → Tensor

Returns a copy of the self tensor. The copy has the same size and data type as self.

contiguous() → Tensor

Returns a contiguous tensor containing the same data as self tensor. If self tensor is contiguous, this function returns the self tensor.

copy_(src, non_blocking=False) → Tensor

Copies the elements from src into self tensor and returns self.

The src tensor must be broadcastable with the self tensor. It may be of a different data type or reside on a different device.

Parameters: src (Tensor) – the source tensor to copy from non_blocking (bool) – if True and this copy is between CPU and GPU, the copy may occur asynchronously with respect to the host. For other cases, this argument has no effect.
cos() → Tensor
cos_() → Tensor

In-place version of cos()

cosh() → Tensor
cosh_() → Tensor

In-place version of cosh()

cpu()
cross(other, dim=-1) → Tensor
cuda(device=None, non_blocking=False) → Tensor

Returns a copy of this object in CUDA memory.

If this object is already in CUDA memory and on the correct device, then no copy is performed and the original object is returned.

Parameters: device (torch.device) – The destination GPU device. Defaults to the current CUDA device. non_blocking (bool) – If True and the source is in pinned memory, the copy will be asynchronous with respect to the host. Otherwise, the argument has no effect. Default: False.
cumprod(dim) → Tensor
cumsum(dim) → Tensor
data_ptr() → int

Returns the address of the first element of self tensor.

det() → Tensor
device
diag(diagonal=0) → Tensor
dim() → int

Returns the number of dimensions of self tensor.

dist(other, p=2) → Tensor
div(value) → Tensor
div_(value) → Tensor

In-place version of div()

dot(tensor2) → Tensor
double() → Tensor

self.double() is equivalent to self.to(torch.float64). See to().

eig(eigenvectors=False) -> (Tensor, Tensor)
element_size() → int

Returns the size in bytes of an individual element.

Example:

>>> torch.tensor([]).element_size()
4
>>> torch.tensor([], dtype=torch.uint8).element_size()
1

eq(other) → Tensor
eq_(other) → Tensor

In-place version of eq()

equal(other) → bool
erf() → Tensor
erf_()
erfinv() → Tensor
erfinv_()
exp() → Tensor
exp_() → Tensor

In-place version of exp()

expm1() → Tensor
expm1_() → Tensor

In-place version of expm1()

expand(*sizes) → Tensor

Returns a new view of the self tensor with singleton dimensions expanded to a larger size.

Passing -1 as the size for a dimension means not changing the size of that dimension.

Tensor can be also expanded to a larger number of dimensions, and the new ones will be appended at the front. For the new dimensions, the size cannot be set to -1.

Expanding a tensor does not allocate new memory, but only creates a new view on the existing tensor where a dimension of size one is expanded to a larger size by setting the stride to 0. Any dimension of size 1 can be expanded to an arbitrary value without allocating new memory.

Parameters: *sizes (torch.Size or int...) – the desired expanded size

Example:

>>> x = torch.tensor([[1], [2], [3]])
>>> x.size()
torch.Size([3, 1])
>>> x.expand(3, 4)
tensor([[ 1,  1,  1,  1],
[ 2,  2,  2,  2],
[ 3,  3,  3,  3]])
>>> x.expand(-1, 4)   # -1 means not changing the size of that dimension
tensor([[ 1,  1,  1,  1],
[ 2,  2,  2,  2],
[ 3,  3,  3,  3]])

expand_as(tensor)[source]
exponential_(lambd=1, *, generator=None) → Tensor

Fills self tensor with elements drawn from the exponential distribution:

$f(x) = \lambda e^{-\lambda x}$
fill_(value) → Tensor

Fills self tensor with the specified value.

float() → Tensor

self.float() is equivalent to self.to(torch.float32). See to().

floor() → Tensor
floor_() → Tensor

In-place version of floor()

fmod(divisor) → Tensor
fmod_(divisor) → Tensor

In-place version of fmod()

frac() → Tensor
frac_() → Tensor

In-place version of frac()

gather(dim, index) → Tensor
ge(other) → Tensor
ge_(other) → Tensor

In-place version of ge()

gels(A) → Tensor
geometric_(p, *, generator=None) → Tensor

Fills self tensor with elements drawn from the geometric distribution:

$f(X=k) = (1 - p)^{k - 1} p$
geqrf() -> (Tensor, Tensor)
ger(vec2) → Tensor
gesv(A) → Tensor, Tensor
gt(other) → Tensor
gt_(other) → Tensor

In-place version of gt()

half() → Tensor

self.half() is equivalent to self.to(torch.float16). See to().

histc(bins=100, min=0, max=0) → Tensor
index(m) → Tensor

Selects elements from self tensor using a binary mask or along a given dimension. The expression tensor.index(m) is equivalent to tensor[m].

Parameters: m (int or ByteTensor or slice) – the dimension or mask used to select elements
index_add_(dim, index, tensor) → Tensor

Accumulate the elements of tensor into the self tensor by adding to the indices in the order given in index. For example, if dim == 0 and index[i] == j, then the ith row of tensor is added to the jth row of self.

The dimth dimension of tensor must have the same size as the length of index (which must be a vector), and all other dimensions must match self, or an error will be raised.

Parameters: dim (int) – dimension along which to index index (LongTensor) – indices of tensor to select from tensor (Tensor) – the tensor containing values to add

Example:

>>> x = torch.ones(5, 3)
>>> t = torch.tensor([[1, 2, 3], [4, 5, 6], [7, 8, 9]], dtype=torch.float)
>>> index = torch.tensor([0, 4, 2])
tensor([[  2.,   3.,   4.],
[  1.,   1.,   1.],
[  8.,   9.,  10.],
[  1.,   1.,   1.],
[  5.,   6.,   7.]])

index_copy_(dim, index, tensor) → Tensor

Copies the elements of tensor into the self tensor by selecting the indices in the order given in index. For example, if dim == 0 and index[i] == j, then the ith row of tensor is copied to the jth row of self.

The dimth dimension of tensor must have the same size as the length of index (which must be a vector), and all other dimensions must match self, or an error will be raised.

Parameters: dim (int) – dimension along which to index index (LongTensor) – indices of tensor to select from tensor (Tensor) – the tensor containing values to copy

Example:

>>> x = torch.zeros(5, 3)
>>> t = torch.tensor([[1, 2, 3], [4, 5, 6], [7, 8, 9]], dtype=torch.float)
>>> index = torch.tensor([0, 4, 2])
>>> x.index_copy_(0, index, t)
tensor([[ 1.,  2.,  3.],
[ 0.,  0.,  0.],
[ 7.,  8.,  9.],
[ 0.,  0.,  0.],
[ 4.,  5.,  6.]])

index_fill_(dim, index, val) → Tensor

Fills the elements of the self tensor with value val by selecting the indices in the order given in index.

Parameters: dim (int) – dimension along which to index index (LongTensor) – indices of self tensor to fill in val (float) – the value to fill with
Example::
>>> x = torch.tensor([[1, 2, 3], [4, 5, 6], [7, 8, 9]], dtype=torch.float)
>>> index = torch.tensor([0, 2])
>>> x.index_fill_(1, index, -1)
tensor([[-1.,  2., -1.],
[-1.,  5., -1.],
[-1.,  8., -1.]])

index_put_(indices, value) → Tensor

Puts values from the tensor value into the tensor self using the indices specified in indices (which is a tuple of Tensors). The expression tensor.index_put_(indices, value) is equivalent to tensor[indices] = value. Returns self.

Parameters: indices (tuple of LongTensor) – tensors used to index into self. value (Tensor) – tensor of same dtype as self.
index_select(dim, index) → Tensor
int() → Tensor

self.int() is equivalent to self.to(torch.int32). See to().

inverse() → Tensor
is_contiguous() → bool

Returns True if self tensor is contiguous in memory in C order.

is_cuda
is_pinned()[source]

Returns true if this tensor resides in pinned memory

is_set_to(tensor) → bool

Returns True if this object refers to the same THTensor object from the Torch C API as the given tensor.

is_signed()
item() → number

Returns the value of this tensor as a standard Python number. This only works for tensors with one element.

This operation is not differentiable.

Example:

>>> x = torch.tensor([1.0])
>>> x.item()
1.0

kthvalue(k, dim=None, keepdim=False) -> (Tensor, LongTensor)
le(other) → Tensor
le_(other) → Tensor

In-place version of le()

lerp(start, end, weight) → Tensor
lerp_(start, end, weight) → Tensor

In-place version of lerp()

log() → Tensor
log_() → Tensor

In-place version of log()

logdet() → Tensor
log10() → Tensor
log10_() → Tensor

In-place version of log10()

log1p() → Tensor
log1p_() → Tensor

In-place version of log1p()

log2() → Tensor
log2_() → Tensor

In-place version of log2()

log_normal_(mean=1, std=2, *, generator=None)

Fills self tensor with numbers samples from the log-normal distribution parameterized by the given mean (µ) and standard deviation (σ). Note that mean and stdv are the mean and standard deviation of the underlying normal distribution, and not of the returned distribution:

$f(x) = \dfrac{1}{x \sigma \sqrt{2\pi}}\ e^{-\dfrac{(\ln x - \mu)^2}{2\sigma^2}}$
long() → Tensor

self.long() is equivalent to self.to(torch.int64). See to().

lt(other) → Tensor
lt_(other) → Tensor

In-place version of lt()

map_(tensor, callable)

Applies callable for each element in self tensor and the given tensor and stores the results in self tensor. self tensor and the given tensor must be broadcastable.

The callable should have the signature:

def callable(a, b) -> number

masked_scatter_(mask, source)

Copies elements from source into self tensor at positions where the mask is one. The shape of mask must be broadcastable with the shape of the underlying tensor. The source should have at least as many elements as the number of ones in mask

Parameters: mask (ByteTensor) – the binary mask source (Tensor) – the tensor to copy from

Note

The mask operates on the self tensor, not on the given source tensor.

masked_fill_(mask, value)

Fills elements of self tensor with value where mask is one. The shape of mask must be broadcastable with the shape of the underlying tensor.

Parameters: mask (ByteTensor) – the binary mask value (float) – the value to fill in with
masked_select(mask) → Tensor
matmul(tensor2) → Tensor
max(dim=None, keepdim=False) -> Tensor or (Tensor, Tensor)
mean(dim=None, keepdim=False) -> Tensor or (Tensor, Tensor)
median(dim=None, keepdim=False) -> (Tensor, LongTensor)
min(dim=None, keepdim=False) -> Tensor or (Tensor, Tensor)
mm(mat2) → Tensor
mode(dim=None, keepdim=False) -> (Tensor, LongTensor)
mul(value) → Tensor
mul_(value)

In-place version of mul()

multinomial(num_samples, replacement=False, *, generator=None) → Tensor
mv(vec) → Tensor
narrow(dimension, start, length) → Tensor

Returns a new tensor that is a narrowed version of self tensor. The dimension dim is narrowed from start to start + length. The returned tensor and self tensor share the same underlying storage.

Parameters: dimension (int) – the dimension along which to narrow start (int) – the starting dimension length (int) – the distance to the ending dimension

Example:

>>> x = torch.tensor([[1, 2, 3], [4, 5, 6], [7, 8, 9]])
>>> x.narrow(0, 0, 2)
tensor([[ 1,  2,  3],
[ 4,  5,  6]])
>>> x.narrow(1, 1, 2)
tensor([[ 2,  3],
[ 5,  6],
[ 8,  9]])

ndimension() → int

Alias for dim()

ne(other) → Tensor
ne_(other) → Tensor

In-place version of ne()

neg() → Tensor
neg_() → Tensor

In-place version of neg()

nelement() → int

Alias for numel()

nonzero() → LongTensor
norm(p=2, dim=None, keepdim=False) → Tensor
normal_(mean=0, std=1, *, generator=None) → Tensor

Fills self tensor with elements samples from the normal distribution parameterized by mean and std.

numel() → int
numpy() → numpy.ndarray

Returns self tensor as a NumPy ndarray. This tensor and the returned ndarray share the same underlying storage. Changes to self tensor will be reflected in the ndarray and vice versa.

orgqr(input2) → Tensor
ormqr(input2, input3, left=True, transpose=False) → Tensor
permute()
pin_memory()
potrf(upper=True) → Tensor
potri(upper=True) → Tensor
potrs(input2, upper=True) → Tensor
pow(exponent) → Tensor
pow_(exponent) → Tensor

In-place version of pow()

prod(dim=None, keepdim=False) → Tensor
pstrf(upper=True, tol=-1) -> (Tensor, IntTensor)
put_(indices, tensor, accumulate=False) → Tensor

Copies the elements from tensor into the positions specified by indices. For the purpose of indexing, the self tensor is treated as if it were a 1-D tensor.

If accumulate is True, the elements in tensor are added to self. If accumulate is False, the behavior is undefined if indices contain duplicate elements.

Parameters: indices (LongTensor) – the indices into self tensor (Tensor) – the tensor containing values to copy from accumulate (bool) – whether to accumulate into self

Example:

>>> src = torch.tensor([[4, 3, 5],
[6, 7, 8]])
>>> src.put_(torch.tensor([1, 3]), torch.tensor([9, 10]))
tensor([[  4,   9,   5],
[ 10,   7,   8]])

qr() -> (Tensor, Tensor)
random_(from=0, to=None, *, generator=None) → Tensor

Fills self tensor with numbers sampled from the discrete uniform distribution over [from, to - 1]. If not specified, the values are usually only bounded by self tensor’s data type. However, for floating point types, if unspecified, range will be [0, 2^mantissa] to ensure that every value is representable. For example, torch.tensor(1, dtype=torch.double).random_() will be uniform in [0, 2^53].

reciprocal() → Tensor
reciprocal_() → Tensor

In-place version of reciprocal()

remainder(divisor) → Tensor
remainder_(divisor) → Tensor

In-place version of remainder()

renorm(p, dim, maxnorm) → Tensor
renorm_(p, dim, maxnorm) → Tensor

In-place version of renorm()

repeat(*sizes) → Tensor

Repeats this tensor along the specified dimensions.

Unlike expand(), this function copies the tensor’s data.

Parameters: sizes (torch.Size or int...) – The number of times to repeat this tensor along each dimension

Example:

>>> x = torch.tensor([1, 2, 3])
>>> x.repeat(4, 2)
tensor([[ 1,  2,  3,  1,  2,  3],
[ 1,  2,  3,  1,  2,  3],
[ 1,  2,  3,  1,  2,  3],
[ 1,  2,  3,  1,  2,  3]])
>>> x.repeat(4, 2, 1).size()
torch.Size([4, 2, 3])

requires_grad_(requires_grad=True) → Tensor

Change if autograd should record operations on this tensor: sets this tensor’s requires_grad attribute in-place. Returns this tensor.

require_grad_()‘s main use case is to tell autograd to begin recording operations on a Tensor tensor. If tensor has requires_grad=False (because it was obtained through a DataLoader, or required preprocessing or initialization), tensor.requires_grad_() makes it so that autograd will begin to record operations on tensor.

Parameters: requires_grad (bool) – If autograd should record operations on this tensor. Default: True.

Example:

>>> # Let's say we want to preprocess some saved weights and use
>>> # the result as new weights.
>>> saved_weights = [0.1, 0.2, 0.3, 0.25]
>>> weights = preprocess(loaded_weights)  # some function
>>> weights
tensor([-0.5503,  0.4926, -2.1158, -0.8303])

>>> # Now, start to record operations done to weights
>>> out = weights.pow(2).sum()
>>> out.backward()
tensor([-1.1007,  0.9853, -4.2316, -1.6606])

reshape(*shape) → Tensor

Returns a tensor with the same data and number of elements as self, but with the specified shape.

Parameters: shape (tuple of python:ints or int...) – the desired shape
resize_(*sizes) → Tensor

Resizes self tensor to the specified size. If the number of elements is larger than the current storage size, then the underlying storage is resized to fit the new number of elements. If the number of elements is smaller, the underlying storage is not changed. Existing elements are preserved but any new memory is uninitialized.

Parameters: sizes (torch.Size or int...) – the desired size

Example:

>>> x = torch.tensor([[1, 2], [3, 4], [5, 6]])
>>> x.resize_(2, 2)
tensor([[ 1,  2],
[ 3,  4]])

resize_as_(tensor) → Tensor

Resizes the self tensor to be the same size as the specified tensor. This is equivalent to self.resize_(tensor.size()).

round() → Tensor
round_() → Tensor

In-place version of round()

rsqrt() → Tensor
rsqrt_() → Tensor

In-place version of rsqrt()

scatter_(dim, index, src) → Tensor

Writes all values from the tensor src into self at the indices specified in the index tensor. For each value in src, its output index is specified by its index in src for dimension != dim and by the corresponding value in index for dimension = dim.

For a 3-D tensor, self is updated as:

self[index[i][j][k]][j][k] = src[i][j][k]  # if dim == 0
self[i][index[i][j][k]][k] = src[i][j][k]  # if dim == 1
self[i][j][index[i][j][k]] = src[i][j][k]  # if dim == 2


This is the reverse operation of the manner described in gather().

self, index and src should have same number of dimensions. It is also required that index->size[d] <= src->size[d] for all dimension d, and that index->size[d] <= real->size[d] for all dimensions d != dim.

Moreover, as for gather(), the values of index must be between 0 and (self.size(dim) -1) inclusive, and all values in a row along the specified dimension dim must be unique.

Parameters: input (Tensor) – the source tensor dim (int) – the axis along which to index index (LongTensor) – the indices of elements to scatter src (Tensor or float) – the source element(s) to scatter

Example:

>>> x = torch.rand(2, 5)
>>> x
tensor([[ 0.3992,  0.2908,  0.9044,  0.4850,  0.6004],
[ 0.5735,  0.9006,  0.6797,  0.4152,  0.1732]])
>>> torch.zeros(3, 5).scatter_(0, torch.tensor([[0, 1, 2, 0, 0], [2, 0, 0, 1, 2]]), x)
tensor([[ 0.3992,  0.9006,  0.6797,  0.4850,  0.6004],
[ 0.0000,  0.2908,  0.0000,  0.4152,  0.0000],
[ 0.5735,  0.0000,  0.9044,  0.0000,  0.1732]])

>>> z = torch.zeros(2, 4).scatter_(1, torch.tensor([[2], [3]]), 1.23)
>>> z
tensor([[ 0.0000,  0.0000,  1.2300,  0.0000],
[ 0.0000,  0.0000,  0.0000,  1.2300]])

select(dim, index) → Tensor

Slices the self tensor along the selected dimension at the given index. This function returns a tensor with the given dimension removed.

Parameters: dim (int) – the dimension to slice index (int) – the index to select with

Note

select() is equivalent to slicing. For example, tensor.select(0, index) is equivalent to tensor[index] and tensor.select(2, index) is equivalent to tensor[:,:,index].

set_(source=None, storage_offset=0, size=None, stride=None) → Tensor

Sets the underlying storage, size, and strides. If source is a tensor, self tensor will share the same storage and have the same size and strides as source. Changes to elements in one tensor will be reflected in the other.

If source is a Storage, the method sets the underlying storage, offset, size, and stride.

Parameters: source (Tensor or Storage) – the tensor or storage to use storage_offset (int, optional) – the offset in the storage size (torch.Size, optional) – the desired size. Defaults to the size of the source. stride (tuple, optional) – the desired stride. Defaults to C-contiguous strides.
share_memory_()[source]

Moves the underlying storage to shared memory.

This is a no-op if the underlying storage is already in shared memory and for CUDA tensors. Tensors in shared memory cannot be resized.

short() → Tensor

self.short() is equivalent to self.to(torch.int16). See to().

sigmoid() → Tensor
sigmoid_() → Tensor

In-place version of sigmoid()

sign() → Tensor
sign_() → Tensor

In-place version of sign()

sin() → Tensor
sin_() → Tensor

In-place version of sin()

sinh() → Tensor
sinh_() → Tensor

In-place version of sinh()

size() → torch.Size

Returns the size of the self tensor. The returned value is a subclass of tuple.

Example:

>>> torch.empty(3, 4, 5).size()
torch.Size([3, 4, 5])

slogdet() -> (Tensor, Tensor)
sort(dim=None, descending=False) -> (Tensor, LongTensor)
split(split_size, dim=0)[source]
sqrt() → Tensor
sqrt_() → Tensor

In-place version of sqrt()

squeeze(dim=None) → Tensor
squeeze_(dim=None) → Tensor

In-place version of squeeze()

std(dim=None, unbiased=True, keepdim=False) → Tensor
storage() → torch.Storage

Returns the underlying storage

storage_offset() → int

Returns self tensor’s offset in the underlying storage in terms of number of storage elements (not bytes).

Example:

>>> x = torch.tensor([1, 2, 3, 4, 5])
>>> x.storage_offset()
0
>>> x[3:].storage_offset()
3

storage_type()
stride(dim) → tuple or int

Returns the stride of self tensor.

Stride is the jump necessary to go from one element to the next one in the specified dimension dim. A tuple of all strides is returned when no argument is passed in. Otherwise, an integer value is returned as the stride in the particular dimension dim.

Parameters: dim (int, optional) – the desired dimension in which stride is required

Example:

>>> x = torch.tensor([[1, 2, 3, 4, 5], [6, 7, 8, 9, 10]])
>>> x.stride()
(5, 1)
>>>x.stride(0)
5
>>> x.stride(-1)
1

sub(value, other) → Tensor

Subtracts a scalar or tensor from self tensor. If both value and other are specified, each element of other is scaled by value before being used.

When other is a tensor, the shape of other must be broadcastable with the shape of the underlying tensor.

sub_(x) → Tensor

In-place version of sub()

sum(dim=None, keepdim=False) → Tensor
svd(some=True) -> (Tensor, Tensor, Tensor)
symeig(eigenvectors=False, upper=True) -> (Tensor, Tensor)
t() → Tensor
t_() → Tensor

In-place version of t()

to(*args, **kwargs) → Tensor

Performs Tensor dtype and/or device conversion. A torch.dtype and torch.device are inferred from the arguments of self.to(*args, **kwargs).

Note

If the self Tensor already has the correct torch.dtype and torch.device, then self is returned. Otherwise, the returned tensor is a copy of self with the desired torch.dtype and torch.device.

Here are the ways to call to:

to(dtype) → Tensor

Returns a Tensor with the specified dtype

to(device, dtype=None) → Tensor

Returns a Tensor with the specified device and (optional) dtype. If dtype is None it is inferred to be self.dtype.

to(other) → Tensor

Returns a Tensor with same torch.dtype and torch.device as the Tensor other.

Example:

>>> tensor = torch.randn(2, 2)  # Initially dtype=float32, device=cpu
>>> tensor.to(torch.float64)
tensor([[-0.5044,  0.0005],
[ 0.3310, -0.0584]], dtype=torch.float64)

>>> cuda0 = torch.device('cuda:0')
>>> tensor.to(cuda0)
tensor([[-0.5044,  0.0005],
[ 0.3310, -0.0584]], device='cuda:0')

>>> tensor.to(cuda0, dtype=torch.float64)
tensor([[-0.5044,  0.0005],
[ 0.3310, -0.0584]], dtype=torch.float64, device='cuda:0')

>>> other = torch.randn((), dtype=torch.float64, device=cuda0)
>>> tensor.to(other)
tensor([[-0.5044,  0.0005],
[ 0.3310, -0.0584]], dtype=torch.float64, device='cuda:0')

take(indices) → Tensor
tan()
tan_() → Tensor

In-place version of tan()

tanh() → Tensor
tanh_() → Tensor

In-place version of tanh()

tolist()
topk(k, dim=None, largest=True, sorted=True) -> (Tensor, LongTensor)
trace() → Tensor
transpose(dim0, dim1) → Tensor
transpose_(dim0, dim1) → Tensor

In-place version of transpose()

tril(k=0) → Tensor
tril_(k=0) → Tensor

In-place version of tril()

triu(k=0) → Tensor
triu_(k=0) → Tensor

In-place version of triu()

trtrs(A, upper=True, transpose=False, unitriangular=False) -> (Tensor, Tensor)
trunc() → Tensor
trunc_() → Tensor

In-place version of trunc()

type(dtype=None, non_blocking=False, **kwargs) → str or Tensor

Returns the type if dtype is not provided, else casts this object to the specified type.

If this is already of the correct type, no copy is performed and the original object is returned.

Parameters: dtype (type or string) – The desired type non_blocking (bool) – If True, and the source is in pinned memory and destination is on the GPU or vice versa, the copy is performed asynchronously with respect to the host. Otherwise, the argument has no effect. **kwargs – For compatibility, may contain the key async in place of the non_blocking argument. The async arg is deprecated.
type_as(tensor) → Tensor

Returns this tensor cast to the type of the given tensor.

This is a no-op if the tensor is already of the correct type. This is equivalent to:

self.type(tensor.type())

Params:
tensor (Tensor): the tensor which has the desired type
unfold(dim, size, step) → Tensor

Returns a tensor which contains all slices of size size from self tensor in the dimension dim.

Step between two slices is given by step.

If sizedim is the size of dimension dim for self, the size of dimension dim in the returned tensor will be (sizedim - size) / step + 1.

An additional dimension of size size is appended in the returned tensor.

Parameters: dim (int) – dimension in which unfolding happens size (int) – the size of each slice that is unfolded step (int) – the step between each slice

Example:

>>> x = torch.arange(1, 8)
>>> x
tensor([ 1.,  2.,  3.,  4.,  5.,  6.,  7.])
>>> x.unfold(0, 2, 1)
tensor([[ 1.,  2.],
[ 2.,  3.],
[ 3.,  4.],
[ 4.,  5.],
[ 5.,  6.],
[ 6.,  7.]])
>>> x.unfold(0, 2, 2)
tensor([[ 1.,  2.],
[ 3.,  4.],
[ 5.,  6.]])

uniform_(from=0, to=1) → Tensor

Fills self tensor with numbers sampled from the continuous uniform distribution:

$P(x) = \dfrac{1}{\text{to} - \text{from}}$
unique(sorted=False, return_inverse=False)[source]

Returns the unique scalar elements of the tensor as a 1-D tensor.

unsqueeze(dim) → Tensor
unsqueeze_(dim) → Tensor

In-place version of unsqueeze()

var(dim=None, unbiased=True, keepdim=False) → Tensor
view(*args) → Tensor

Returns a new tensor with the same data as the self tensor but of a different size.

The returned tensor shares the same data and must have the same number of elements, but may have a different size. For a tensor to be viewed, the new view size must be compatible with its original size and stride, i.e., each new view dimension must either be a subspace of an original dimension, or only span across original dimensions $$d, d+1, \dots, d+k$$ that satisfy the following contiguity-like condition that $$\forall i = 0, \dots, k-1$$,

$stride[i] = stride[i+1] \times size[i+1]$

Otherwise, contiguous() needs to be called before the tensor can be viewed.

Parameters: args (torch.Size or int...) – the desired size

Example:

>>> x = torch.randn(4, 4)
>>> x.size()
torch.Size([4, 4])
>>> y = x.view(16)
>>> y.size()
torch.Size([16])
>>> z = x.view(-1, 8)  # the size -1 is inferred from other dimensions
>>> z.size()
torch.Size([2, 8])

view_as(other) → Tensor[source]

View this tensor as the same size as other. self.view_as(other) is equivalent to self.view(other.size()).

Parameters: other (torch.Tensor) – The result tensor has the same size as other.size().
zero_() → Tensor

Fills self tensor with zeros.

class torch.ByteTensor

The following methods are unique to torch.ByteTensor.

all() → bool

Returns True if all elements in the tensor are non-zero, False otherwise.

any() → bool

Returns True if any elements in the tensor are non-zero, False otherwise.