Shortcuts

torch.Tensor

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

Torch defines nine CPU tensor types and nine 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

Boolean

torch.bool

torch.BoolTensor

torch.cuda.BoolTensor

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.as_tensor().

A 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()
>>> x.grad
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

For more information on the torch.dtype, torch.device, and torch.layout attributes of a torch.Tensor, see Tensor Attributes.

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.

Warning

Current implementation of torch.Tensor introduces memory overhead, thus it might lead to unexpectedly high memory usage in the applications with many tiny tensors. If this is your case, consider using one large structure.

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().

Warning

When data is a tensor x, new_tensor() reads out ‘the data’ from whatever it is passed, and constructs a leaf variable. Therefore tensor.new_tensor(x) is equivalent to x.clone().detach() and tensor.new_tensor(x, requires_grad=True) is equivalent to x.clone().detach().requires_grad_(True). The equivalents using clone() and detach() are recommended.

Parameters
  • data (array_like) – The returned Tensor copies data.

  • dtype (torch.dtype, optional) – the desired type of returned tensor. Default: if None, same torch.dtype as this tensor.

  • device (torch.device, optional) – the desired device of returned tensor. Default: if None, same torch.device as this tensor.

  • requires_grad (bool, optional) – If autograd should record operations on the returned tensor. Default: False.

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. Default: if None, same torch.dtype as this tensor.

  • device (torch.device, optional) – the desired device of returned tensor. Default: if None, same torch.device as this tensor.

  • requires_grad (bool, optional) – If autograd should record operations on the returned tensor. Default: False.

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. Default: if None, same torch.dtype as this tensor.

  • device (torch.device, optional) – the desired device of returned tensor. Default: if None, same torch.device as this tensor.

  • requires_grad (bool, optional) – If autograd should record operations on the returned tensor. Default: False.

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. Default: if None, same torch.dtype as this tensor.

  • device (torch.device, optional) – the desired device of returned tensor. Default: if None, same torch.device as this tensor.

  • requires_grad (bool, optional) – If autograd should record operations on the returned tensor. Default: False.

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. Default: if None, same torch.dtype as this tensor.

  • device (torch.device, optional) – the desired device of returned tensor. Default: if None, same torch.device as this tensor.

  • requires_grad (bool, optional) – If autograd should record operations on the returned tensor. Default: False.

Example:

>>> tensor = torch.tensor((), dtype=torch.float64)
>>> tensor.new_zeros((2, 3))
tensor([[ 0.,  0.,  0.],
        [ 0.,  0.,  0.]], dtype=torch.float64)
is_cuda

Is True if the Tensor is stored on the GPU, False otherwise.

device

Is the torch.device where this Tensor is.

grad

This attribute is None by default and becomes a Tensor the first time a call to backward() computes gradients for self. The attribute will then contain the gradients computed and future calls to backward() will accumulate (add) gradients into it.

ndim

Alias for dim()

T

Is this Tensor with its dimensions reversed.

If n is the number of dimensions in x, x.T is equivalent to x.permute(n-1, n-2, ..., 0).

abs() → Tensor

See torch.abs()

abs_() → Tensor

In-place version of abs()

acos() → Tensor

See torch.acos()

acos_() → Tensor

In-place version of acos()

add(value) → Tensor

add(value=1, other) -> Tensor

See torch.add()

add_(value) → Tensor

add_(value=1, other) -> Tensor

In-place version of add()

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

See torch.addbmm()

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

In-place version of addbmm()

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

See torch.addcdiv()

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

In-place version of addcdiv()

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

See torch.addcmul()

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

In-place version of addcmul()

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

See torch.addmm()

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

In-place version of addmm()

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

See torch.addmv()

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

In-place version of addmv()

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

See torch.addr()

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

In-place version of addr()

allclose(other, rtol=1e-05, atol=1e-08, equal_nan=False) → Tensor

See torch.allclose()

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) → LongTensor

See torch.argmax()

argmin(dim=None, keepdim=False) → LongTensor

See torch.argmin()

argsort(dim=-1, descending=False) → LongTensor

See :func: torch.argsort

asin() → Tensor

See torch.asin()

asin_() → Tensor

In-place version of asin()

as_strided(size, stride, storage_offset=0) → Tensor

See torch.as_strided()

atan() → Tensor

See torch.atan()

atan2(other) → Tensor

See torch.atan2()

atan2_(other) → Tensor

In-place version of atan2()

atan_() → Tensor

In-place version of atan()

backward(gradient=None, retain_graph=None, create_graph=False)[source]

Computes the gradient of current tensor w.r.t. graph leaves.

The graph is differentiated using the chain rule. If the tensor is non-scalar (i.e. its data has more than one element) and requires gradient, the function additionally requires specifying gradient. It should be a tensor of matching type and location, that contains the gradient of the differentiated function w.r.t. self.

This function accumulates gradients in the leaves - you might need to zero them before calling it.

Parameters
  • gradient (Tensor or None) – Gradient w.r.t. the tensor. If it is a tensor, it will be automatically converted to a Tensor that does not require grad unless create_graph is True. None values can be specified for scalar Tensors or ones that don’t require grad. If a None value would be acceptable then this argument is optional.

  • retain_graph (bool, optional) – If False, the graph used to compute the grads will be freed. Note that in nearly all cases setting this option to True is not needed and often can be worked around in a much more efficient way. Defaults to the value of create_graph.

  • create_graph (bool, optional) – If True, graph of the derivative will be constructed, allowing to compute higher order derivative products. Defaults to False.

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

See torch.baddbmm()

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

In-place version of baddbmm()

bernoulli(*, generator=None) → Tensor

Returns a result tensor where each result[i]\texttt{result[i]} is independently sampled from Bernoulli(self[i])\text{Bernoulli}(\texttt{self[i]}) . self must have floating point dtype, and the result will have the same dtype.

See torch.bernoulli()

bernoulli_()
bernoulli_(p=0.5, *, generator=None) → Tensor

Fills each location of self with an independent sample from Bernoulli(p)\text{Bernoulli}(\texttt{p}) . self can have integral dtype.

bernoulli_(p_tensor, *, generator=None) → Tensor

p_tensor should be a tensor containing probabilities to be used for drawing the binary random number.

The ith\text{i}^{th} element of self tensor will be set to a value sampled from Bernoulli(p_tensor[i])\text{Bernoulli}(\texttt{p\_tensor[i]}) .

self can have integral dtype, but p_tensor must have floating point dtype.

See also bernoulli() and torch.bernoulli()

bfloat16() → Tensor

self.bfloat16() is equivalent to self.to(torch.bfloat16). See to().

bincount(weights=None, minlength=0) → Tensor

See torch.bincount()

bitwise_not() → Tensor

See torch.bitwise_not()

bitwise_not_() → Tensor

In-place version of bitwise_not()

bmm(batch2) → Tensor

See torch.bmm()

bool() → Tensor

self.bool() is equivalent to self.to(torch.bool). See to().

byte() → Tensor

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

cauchy_(median=0, sigma=1, *, generator=None) → Tensor

Fills the tensor with numbers drawn from the Cauchy distribution:

f(x)=1πσ(xmedian)2+σ2f(x) = \dfrac{1}{\pi} \dfrac{\sigma}{(x - \text{median})^2 + \sigma^2}
ceil() → Tensor

See torch.ceil()

ceil_() → Tensor

In-place version of ceil()

char() → Tensor

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

cholesky(upper=False) → Tensor

See torch.cholesky()

cholesky_inverse(upper=False) → Tensor

See torch.cholesky_inverse()

cholesky_solve(input2, upper=False) → Tensor

See torch.cholesky_solve()

chunk(chunks, dim=0) → List of Tensors

See torch.chunk()

clamp(min, max) → Tensor

See torch.clamp()

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.

Note

Unlike copy_(), this function is recorded in the computation graph. Gradients propagating to the cloned tensor will propagate to the original tensor.

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

See torch.cos()

cos_() → Tensor

In-place version of cos()

cosh() → Tensor

See torch.cosh()

cosh_() → Tensor

In-place version of cosh()

cpu() → Tensor

Returns a copy of this object in CPU memory.

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

cross(other, dim=-1) → Tensor

See torch.cross()

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, dtype=None) → Tensor

See torch.cumprod()

cumsum(dim, dtype=None) → Tensor

See torch.cumsum()

data_ptr() → int

Returns the address of the first element of self tensor.

dequantize() → Tensor

Given a quantized Tensor, dequantize it and return the dequantized float Tensor.

det() → Tensor

See torch.det()

dense_dim() → int

If self is a sparse COO tensor (i.e., with torch.sparse_coo layout), this returns the number of dense dimensions. Otherwise, this throws an error.

See also Tensor.sparse_dim().

detach()

Returns a new Tensor, detached from the current graph.

The result will never require gradient.

Note

Returned Tensor shares the same storage with the original one. In-place modifications on either of them will be seen, and may trigger errors in correctness checks. IMPORTANT NOTE: Previously, in-place size / stride / storage changes (such as resize_ / resize_as_ / set_ / transpose_) to the returned tensor also update the original tensor. Now, these in-place changes will not update the original tensor anymore, and will instead trigger an error. For sparse tensors: In-place indices / values changes (such as zero_ / copy_ / add_) to the returned tensor will not update the original tensor anymore, and will instead trigger an error.

detach_()

Detaches the Tensor from the graph that created it, making it a leaf. Views cannot be detached in-place.

diag(diagonal=0) → Tensor

See torch.diag()

diag_embed(offset=0, dim1=-2, dim2=-1) → Tensor

See torch.diag_embed()

diagflat(offset=0) → Tensor

See torch.diagflat()

diagonal(offset=0, dim1=0, dim2=1) → Tensor

See torch.diagonal()

fill_diagonal_(fill_value, wrap=False) → Tensor

Fill the main diagonal of a tensor that has at least 2-dimensions. When dims>2, all dimensions of input must be of equal length. This function modifies the input tensor in-place, and returns the input tensor.

Parameters
  • fill_value (Scalar) – the fill value

  • wrap (bool) – the diagonal ‘wrapped’ after N columns for tall matrices.

Example:

>>> a = torch.zeros(3, 3)
>>> a.fill_diagonal_(5)
tensor([[5., 0., 0.],
        [0., 5., 0.],
        [0., 0., 5.]])
>>> b = torch.zeros(7, 3)
>>> b.fill_diagonal_(5)
tensor([[5., 0., 0.],
        [0., 5., 0.],
        [0., 0., 5.],
        [0., 0., 0.],
        [0., 0., 0.],
        [0., 0., 0.],
        [0., 0., 0.]])
>>> c = torch.zeros(7, 3)
>>> c.fill_diagonal_(5, wrap=True)
tensor([[5., 0., 0.],
        [0., 5., 0.],
        [0., 0., 5.],
        [0., 0., 0.],
        [5., 0., 0.],
        [0., 5., 0.],
        [0., 0., 5.]])
digamma() → Tensor

See torch.digamma()

digamma_() → Tensor

In-place version of digamma()

dim() → int

Returns the number of dimensions of self tensor.

dist(other, p=2) → Tensor

See torch.dist()

div(value) → Tensor

See torch.div()

div_(value) → Tensor

In-place version of div()

dot(tensor2) → Tensor

See torch.dot()

double() → Tensor

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

eig(eigenvectors=False) -> (Tensor, Tensor)

See torch.eig()

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

See torch.eq()

eq_(other) → Tensor

In-place version of eq()

equal(other) → bool

See torch.equal()

erf() → Tensor

See torch.erf()

erf_() → Tensor

In-place version of erf()

erfc() → Tensor

See torch.erfc()

erfc_() → Tensor

In-place version of erfc()

erfinv() → Tensor

See torch.erfinv()

erfinv_() → Tensor

In-place version of erfinv()

exp() → Tensor

See torch.exp()

exp_() → Tensor

In-place version of exp()

expm1() → Tensor

See torch.expm1()

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

Warning

More than one element of an expanded tensor may refer to a single memory location. As a result, in-place operations (especially ones that are vectorized) may result in incorrect behavior. If you need to write to the tensors, please clone them first.

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(other) → Tensor

Expand this tensor to the same size as other. self.expand_as(other) is equivalent to self.expand(other.size()).

Please see expand() for more information about expand.

Parameters

other (torch.Tensor) – The result tensor has the same size as other.

exponential_(lambd=1, *, generator=None) → Tensor

Fills self tensor with elements drawn from the exponential distribution:

f(x)=λeλxf(x) = \lambda e^{-\lambda x}
fft(signal_ndim, normalized=False) → Tensor

See torch.fft()

fill_(value) → Tensor

Fills self tensor with the specified value.

flatten(input, start_dim=0, end_dim=-1) → Tensor

see torch.flatten()

flip(dims) → Tensor

See torch.flip()

float() → Tensor

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

floor() → Tensor

See torch.floor()

floor_() → Tensor

In-place version of floor()

fmod(divisor) → Tensor

See torch.fmod()

fmod_(divisor) → Tensor

In-place version of fmod()

frac() → Tensor

See torch.frac()

frac_() → Tensor

In-place version of frac()

gather(dim, index) → Tensor

See torch.gather()

ge(other) → Tensor

See torch.ge()

ge_(other) → Tensor

In-place version of ge()

gels(A)[source]

See torch.lstsq()

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

Fills self tensor with elements drawn from the geometric distribution:

f(X=k)=pk1(1p)f(X=k) = p^{k - 1} (1 - p)
geqrf() -> (Tensor, Tensor)

See torch.geqrf()

ger(vec2) → Tensor

See torch.ger()

get_device() -> Device ordinal (Integer)

For CUDA tensors, this function returns the device ordinal of the GPU on which the tensor resides. For CPU tensors, an error is thrown.

Example:

>>> x = torch.randn(3, 4, 5, device='cuda:0')
>>> x.get_device()
0
>>> x.cpu().get_device()  # RuntimeError: get_device is not implemented for type torch.FloatTensor
gt(other) → Tensor

See torch.gt()

gt_(other) → Tensor

In-place version of gt()

half() → Tensor

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

hardshrink(lambd=0.5) → Tensor

See torch.nn.functional.hardshrink()

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

See torch.histc()

ifft(signal_ndim, normalized=False) → Tensor

See torch.ifft()

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.

Note

When using the CUDA backend, this operation may induce nondeterministic behaviour that is not easily switched off. Please see the notes on Reproducibility for background.

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])
>>> x.index_add_(0, index, t)
tensor([[  2.,   3.,   4.],
        [  1.,   1.,   1.],
        [  8.,   9.,  10.],
        [  1.,   1.,   1.],
        [  5.,   6.,   7.]])
index_add(dim, index, tensor) → Tensor

Out-of-place version of torch.Tensor.index_add_()

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_copy(dim, index, tensor) → Tensor

Out-of-place version of torch.Tensor.index_copy_()

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_fill(dim, index, value) → Tensor

Out-of-place version of torch.Tensor.index_fill_()

index_put_(indices, value, accumulate=False) → 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.

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 (tuple of LongTensor) – tensors used to index into self.

  • value (Tensor) – tensor of same dtype as self.

  • accumulate (bool) – whether to accumulate into self

index_put(indices, value, accumulate=False) → Tensor

Out-place version of index_put_()

index_select(dim, index) → Tensor

See torch.index_select()

indices() → Tensor

If self is a sparse COO tensor (i.e., with torch.sparse_coo layout), this returns a view of the contained indices tensor. Otherwise, this throws an error.

See also Tensor.values().

Note

This method can only be called on a coalesced sparse tensor. See Tensor.coalesce() for details.

int() → Tensor

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

int_repr() → Tensor

Given a quantized Tensor, self.int_repr() returns a CPU Tensor with uint8_t as data type that stores the underlying uint8_t values of the given Tensor.

inverse() → Tensor

See torch.inverse()

irfft(signal_ndim, normalized=False, onesided=True, signal_sizes=None) → Tensor

See torch.irfft()

is_contiguous() → bool

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

is_floating_point() → bool

Returns True if the data type of self is a floating point data type.

is_leaf()

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 None.

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:

>>> a = torch.rand(10, requires_grad=True)
>>> a.is_leaf
True
>>> b = torch.rand(10, requires_grad=True).cuda()
>>> b.is_leaf
False
# b was created by the operation that cast a cpu Tensor into a cuda Tensor
>>> c = torch.rand(10, requires_grad=True) + 2
>>> c.is_leaf
False
# c was created by the addition operation
>>> d = torch.rand(10).cuda()
>>> d.is_leaf
True
# d does not require gradients and so has no operation creating it (that is tracked by the autograd engine)
>>> e = torch.rand(10).cuda().requires_grad_()
>>> e.is_leaf
True
# e requires gradients and has no operations creating it
>>> f = torch.rand(10, requires_grad=True, device="cuda")
>>> f.is_leaf
True
# f requires grad, has no operation creating it
is_pinned()

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_shared()[source]

Checks if tensor is in shared memory.

This is always True for CUDA tensors.

is_signed() → bool

Returns True if the data type of self is a signed data type.

is_sparse()
item() → number

Returns the value of this tensor as a standard Python number. This only works for tensors with one element. For other cases, see tolist().

This operation is not differentiable.

Example:

>>> x = torch.tensor([1.0])
>>> x.item()
1.0
kthvalue(k, dim=None, keepdim=False) -> (Tensor, LongTensor)

See torch.kthvalue()

le(other) → Tensor

See torch.le()

le_(other) → Tensor

In-place version of le()

lerp(end, weight) → Tensor

See torch.lerp()

lerp_(end, weight) → Tensor

In-place version of lerp()

log() → Tensor

See torch.log()

log_() → Tensor

In-place version of log()

logdet() → Tensor

See torch.logdet()

log10() → Tensor

See torch.log10()

log10_() → Tensor

In-place version of log10()

log1p() → Tensor

See torch.log1p()

log1p_() → Tensor

In-place version of log1p()

log2() → Tensor

See torch.log2()

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\mu and standard deviation σ\sigma . Note that mean and std are the mean and standard deviation of the underlying normal distribution, and not of the returned distribution:

f(x)=1xσ2π e(lnxμ)22σ2f(x) = \dfrac{1}{x \sigma \sqrt{2\pi}}\ e^{-\frac{(\ln x - \mu)^2}{2\sigma^2}}
logsumexp(dim, keepdim=False) → Tensor

See torch.logsumexp()

logical_not() → Tensor

See torch.logical_not()

logical_not_() → Tensor

In-place version of logical_not()

logical_xor() → Tensor

See torch.logical_xor()

logical_xor_() → Tensor

In-place version of logical_xor()

long() → Tensor

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

lstsq(A) -> (Tensor, Tensor)

See torch.lstsq()

lt(other) → Tensor

See torch.lt()

lt_(other) → Tensor

In-place version of lt()

lu(pivot=True, get_infos=False)[source]

See torch.lu()

lu_solve(LU_data, LU_pivots) → Tensor

See torch.lu_solve()

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 True. 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 (BoolTensor) – the boolean mask

  • source (Tensor) – the tensor to copy from

Note

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

masked_scatter(mask, tensor) → Tensor

Out-of-place version of torch.Tensor.masked_scatter_()

masked_fill_(mask, value)

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

Parameters
  • mask (BoolTensor) – the boolean mask

  • value (float) – the value to fill in with

masked_fill(mask, value) → Tensor

Out-of-place version of torch.Tensor.masked_fill_()

masked_select(mask) → Tensor

See torch.masked_select()

matmul(tensor2) → Tensor

See torch.matmul()

matrix_power(n) → Tensor

See torch.matrix_power()

max(dim=None, keepdim=False) -> Tensor or (Tensor, Tensor)

See torch.max()

mean(dim=None, keepdim=False) -> Tensor or (Tensor, Tensor)

See torch.mean()

median(dim=None, keepdim=False) -> (Tensor, LongTensor)

See torch.median()

min(dim=None, keepdim=False) -> Tensor or (Tensor, Tensor)

See torch.min()

mm(mat2) → Tensor

See torch.mm()

mode(dim=None, keepdim=False) -> (Tensor, LongTensor)

See torch.mode()

mul(value) → Tensor

See torch.mul()

mul_(value)

In-place version of mul()

multinomial(num_samples, replacement=False, *, generator=None) → Tensor

See torch.multinomial()

mv(vec) → Tensor

See torch.mv()

mvlgamma(p) → Tensor

See torch.mvlgamma()

mvlgamma_(p) → Tensor

In-place version of mvlgamma()

narrow(dimension, start, length) → Tensor

See torch.narrow()

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]])
narrow_copy(dimension, start, length) → Tensor

Same as Tensor.narrow() except returning a copy rather than shared storage. This is primarily for sparse tensors, which do not have a shared-storage narrow method. Calling `narrow_copy with `dimemsion > self.sparse_dim()` will return a copy with the relevant dense dimension narrowed, and `self.shape` updated accordingly.

ndimension() → int

Alias for dim()

ne(other) → Tensor

See torch.ne()

ne_(other) → Tensor

In-place version of ne()

neg() → Tensor

See torch.neg()

neg_() → Tensor

In-place version of neg()

nelement() → int

Alias for numel()

nonzero() → LongTensor

See torch.nonzero()

norm(p='fro', dim=None, keepdim=False, dtype=None)[source]

See torch.norm()

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

See torch.numel()

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

See torch.orgqr()

ormqr(input2, input3, left=True, transpose=False) → Tensor

See torch.ormqr()

permute(*dims) → Tensor

Permute the dimensions of this tensor.

Parameters

*dims (int...) – The desired ordering of dimensions

Example

>>> x = torch.randn(2, 3, 5)
>>> x.size()
torch.Size([2, 3, 5])
>>> x.permute(2, 0, 1).size()
torch.Size([5, 2, 3])
pin_memory() → Tensor

Copies the tensor to pinned memory, if it’s not already pinned.

pinverse() → Tensor

See torch.pinverse()

pow(exponent) → Tensor

See torch.pow()

pow_(exponent) → Tensor

In-place version of pow()

prod(dim=None, keepdim=False, dtype=None) → Tensor

See torch.prod()

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(some=True) -> (Tensor, Tensor)

See torch.qr()

qscheme() → torch.qscheme

Returns the quantization scheme of a given QTensor.

q_scale() → float

Given a Tensor quantized by linear(affine) quantization, returns the scale of the underlying quantizer().

q_zero_point() → int

Given a Tensor quantized by linear(affine) quantization, returns the zero_point of the underlying quantizer().

q_per_channel_scales() → Tensor

Given a Tensor quantized by linear (affine) per-channel quantization, returns a Tensor of scales of the underlying quantizer().

q_per_channel_zero_points() → Tensor

Given a Tensor quantized by linear (affine) per-channel quantization, returns a tensor of zero_points of the underlying quantizer().

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

See torch.reciprocal()

reciprocal_() → Tensor

In-place version of reciprocal()

record_stream(stream)

Ensures that the tensor memory is not reused for another tensor until all current work queued on stream are complete.

Note

The caching allocator is aware of only the stream where a tensor was allocated. Due to the awareness, it already correctly manages the life cycle of tensors on only one stream. But if a tensor is used on a stream different from the stream of origin, the allocator might reuse the memory unexpectedly. Calling this method lets the allocator know which streams have used the tensor.

register_hook(hook)[source]

Registers a backward hook.

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

hook(grad) -> Tensor or None

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 a handle with a method handle.remove() that removes the hook from the module.

Example:

>>> v = torch.tensor([0., 0., 0.], requires_grad=True)
>>> h = v.register_hook(lambda grad: grad * 2)  # double the gradient
>>> v.backward(torch.tensor([1., 2., 3.]))
>>> v.grad

 2
 4
 6
[torch.FloatTensor of size (3,)]

>>> h.remove()  # removes the hook
remainder(divisor) → Tensor

See torch.remainder()

remainder_(divisor) → Tensor

In-place version of remainder()

renorm(p, dim, maxnorm) → Tensor

See torch.renorm()

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.

Warning

torch.repeat() behaves differently from numpy.repeat, but is more similar to numpy.tile. For the operator similar to numpy.repeat, see torch.repeat_interleave().

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])
repeat_interleave(repeats, dim=None) → Tensor

See torch.repeat_interleave().

requires_grad()

Is True if gradients need to be computed for this Tensor, False otherwise.

Note

The fact that gradients need to be computed for a Tensor do not mean that the grad attribute will be populated, see is_leaf for more details.

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.

requires_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]
>>> loaded_weights = torch.tensor(saved_weights)
>>> weights = preprocess(loaded_weights)  # some function
>>> weights
tensor([-0.5503,  0.4926, -2.1158, -0.8303])

>>> # Now, start to record operations done to weights
>>> weights.requires_grad_()
>>> out = weights.pow(2).sum()
>>> out.backward()
>>> weights.grad
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. This method returns a view if shape is compatible with the current shape. See torch.Tensor.view() on when it is possible to return a view.

See torch.reshape()

Parameters

shape (tuple of python:ints or int...) – the desired shape

reshape_as(other) → Tensor

Returns this tensor as the same shape as other. self.reshape_as(other) is equivalent to self.reshape(other.sizes()). This method returns a view if other.sizes() is compatible with the current shape. See torch.Tensor.view() on when it is possible to return a view.

Please see reshape() for more information about reshape.

Parameters

other (torch.Tensor) – The result tensor has the same shape as other.

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.

Warning

This is a low-level method. The storage is reinterpreted as C-contiguous, ignoring the current strides (unless the target size equals the current size, in which case the tensor is left unchanged). For most purposes, you will instead want to use view(), which checks for contiguity, or reshape(), which copies data if needed. To change the size in-place with custom strides, see set_().

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()).

retain_grad()[source]

Enables .grad attribute for non-leaf Tensors.

rfft(signal_ndim, normalized=False, onesided=True) → Tensor

See torch.rfft()

roll(shifts, dims) → Tensor

See torch.roll()

rot90(k, dims) → Tensor

See torch.rot90()

round() → Tensor

See torch.round()

round_() → Tensor

In-place version of round()

rsqrt() → Tensor

See torch.rsqrt()

rsqrt_() → Tensor

In-place version of rsqrt()

scatter(dim, index, source) → Tensor

Out-of-place version of torch.Tensor.scatter_()

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 (if it is a Tensor) should have same number of dimensions. It is also required that index.size(d) <= src.size(d) for all dimensions d, and that index.size(d) <= self.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
  • dim (int) – the axis along which to index

  • index (LongTensor) – the indices of elements to scatter, can be either empty or the same size of src. When empty, the operation returns identity

  • src (Tensor) – the source element(s) to scatter, incase value is not specified

  • value (float) – the source element(s) to scatter, incase src is not specified

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]])
scatter_add_(dim, index, other) → Tensor

Adds all values from the tensor other into self at the indices specified in the index tensor in a similar fashion as scatter_(). For each value in other, it is added to an index in self which is specified by its index in other 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] += other[i][j][k]  # if dim == 0
self[i][index[i][j][k]][k] += other[i][j][k]  # if dim == 1
self[i][j][index[i][j][k]] += other[i][j][k]  # if dim == 2

self, index and other should have same number of dimensions. It is also required that index.size(d) <= other.size(d) for all dimensions d, and that index.size(d) <= self.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.

Note

When using the CUDA backend, this operation may induce nondeterministic behaviour that is not easily switched off. Please see the notes on Reproducibility for background.

Parameters
  • dim (int) – the axis along which to index

  • index (LongTensor) – the indices of elements to scatter and add, can be either empty or the same size of src. When empty, the operation returns identity.

  • other (Tensor) – the source elements to scatter and add

Example:

>>> x = torch.rand(2, 5)
>>> x
tensor([[0.7404, 0.0427, 0.6480, 0.3806, 0.8328],
        [0.7953, 0.2009, 0.9154, 0.6782, 0.9620]])
>>> torch.ones(3, 5).scatter_add_(0, torch.tensor([[0, 1, 2, 0, 0], [2, 0, 0, 1, 2]]), x)
tensor([[1.7404, 1.2009, 1.9154, 1.3806, 1.8328],
        [1.0000, 1.0427, 1.0000, 1.6782, 1.0000],
        [1.7953, 1.0000, 1.6480, 1.0000, 1.9620]])
scatter_add(dim, index, source) → Tensor

Out-of-place version of torch.Tensor.scatter_add_()

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

See torch.sigmoid()

sigmoid_() → Tensor

In-place version of sigmoid()

sign() → Tensor

See torch.sign()

sign_() → Tensor

In-place version of sign()

sin() → Tensor

See torch.sin()

sin_() → Tensor

In-place version of sin()

sinh() → Tensor

See torch.sinh()

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)

See torch.slogdet()

solve(A) → Tensor, Tensor

See torch.solve()

sort(dim=-1, descending=False) -> (Tensor, LongTensor)

See torch.sort()

split(split_size, dim=0)[source]

See torch.split()

sparse_mask(input, mask) → Tensor

Returns a new SparseTensor with values from Tensor input filtered by indices of mask and values are ignored. input and mask must have the same shape.

Parameters
  • input (Tensor) – an input Tensor

  • mask (SparseTensor) – a SparseTensor which we filter input based on its indices

Example:

>>> nnz = 5
>>> dims = [5, 5, 2, 2]
>>> I = torch.cat([torch.randint(0, dims[0], size=(nnz,)),
                   torch.randint(0, dims[1], size=(nnz,))], 0).reshape(2, nnz)
>>> V = torch.randn(nnz, dims[2], dims[3])
>>> size = torch.Size(dims)
>>> S = torch.sparse_coo_tensor(I, V, size).coalesce()
>>> D = torch.randn(dims)
>>> D.sparse_mask(S)
tensor(indices=tensor([[0, 0, 0, 2],
                       [0, 1, 4, 3]]),
       values=tensor([[[ 1.6550,  0.2397],
                       [-0.1611, -0.0779]],

                      [[ 0.2326, -1.0558],
                       [ 1.4711,  1.9678]],

                      [[-0.5138, -0.0411],
                       [ 1.9417,  0.5158]],

                      [[ 0.0793,  0.0036],
                       [-0.2569, -0.1055]]]),
       size=(5, 5, 2, 2), nnz=4, layout=torch.sparse_coo)
sparse_dim() → int

If self is a sparse COO tensor (i.e., with torch.sparse_coo layout), this returns the number of sparse dimensions. Otherwise, this throws an error.

See also Tensor.dense_dim().

sqrt() → Tensor

See torch.sqrt()

sqrt_() → Tensor

In-place version of sqrt()

squeeze(dim=None) → Tensor

See torch.squeeze()

squeeze_(dim=None) → Tensor

In-place version of squeeze()

std(dim=None, unbiased=True, keepdim=False) → Tensor

See torch.std()

stft(n_fft, hop_length=None, win_length=None, window=None, center=True, pad_mode='reflect', normalized=False, onesided=True)[source]

See torch.stft()

Warning

This function changed signature at version 0.4.1. Calling with the previous signature may cause error or return incorrect result.

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() → type

Returns the type of the underlying storage.

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, dtype=None) → Tensor

See torch.sum()

sum_to_size(*size) → Tensor

Sum this tensor to size. size must be broadcastable to this tensor size.

Parameters

size (int...) – a sequence of integers defining the shape of the output tensor.

svd(some=True, compute_uv=True) -> (Tensor, Tensor, Tensor)

See torch.svd()

symeig(eigenvectors=False, upper=True) -> (Tensor, Tensor)

See torch.symeig()

t() → Tensor

See torch.t()

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, non_blocking=False, copy=False) → Tensor

Returns a Tensor with the specified dtype

to(device=None, dtype=None, non_blocking=False, copy=False) → Tensor

Returns a Tensor with the specified device and (optional) dtype. If dtype is None it is inferred to be self.dtype. When non_blocking, tries to convert asynchronously with respect to the host if possible, e.g., converting a CPU Tensor with pinned memory to a CUDA Tensor. When copy is set, a new Tensor is created even when the Tensor already matches the desired conversion.

to(other, non_blocking=False, copy=False) → Tensor

Returns a Tensor with same torch.dtype and torch.device as the Tensor other. When non_blocking, tries to convert asynchronously with respect to the host if possible, e.g., converting a CPU Tensor with pinned memory to a CUDA Tensor. When copy is set, a new Tensor is created even when the Tensor already matches the desired conversion.

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, non_blocking=True)
tensor([[-0.5044,  0.0005],
        [ 0.3310, -0.0584]], dtype=torch.float64, device='cuda:0')
to_mkldnn() → Tensor

Returns a copy of the tensor in torch.mkldnn layout.

take(indices) → Tensor

See torch.take()

tan() → Tensor

See torch.tan()

tan_() → Tensor

In-place version of tan()

tanh() → Tensor

See torch.tanh()

tanh_() → Tensor

In-place version of tanh()

tolist()

” tolist() -> list or number

Returns the tensor as a (nested) list. For scalars, a standard Python number is returned, just like with item(). Tensors are automatically moved to the CPU first if necessary.

This operation is not differentiable.

Examples:

>>> a = torch.randn(2, 2)
>>> a.tolist()
[[0.012766935862600803, 0.5415473580360413],
 [-0.08909505605697632, 0.7729271650314331]]
>>> a[0,0].tolist()
0.012766935862600803
topk(k, dim=None, largest=True, sorted=True) -> (Tensor, LongTensor)

See torch.topk()

to_sparse(sparseDims) → Tensor

Returns a sparse copy of the tensor. PyTorch supports sparse tensors in coordinate format.

Parameters

sparseDims (int, optional) – the number of sparse dimensions to include in the new sparse tensor

Example:

>>> d = torch.tensor([[0, 0, 0], [9, 0, 10], [0, 0, 0]])
>>> d
tensor([[ 0,  0,  0],
        [ 9,  0, 10],
        [ 0,  0,  0]])
>>> d.to_sparse()
tensor(indices=tensor([[1, 1],
                       [0, 2]]),
       values=tensor([ 9, 10]),
       size=(3, 3), nnz=2, layout=torch.sparse_coo)
>>> d.to_sparse(1)
tensor(indices=tensor([[1]]),
       values=tensor([[ 9,  0, 10]]),
       size=(3, 3), nnz=1, layout=torch.sparse_coo)
trace() → Tensor

See torch.trace()

transpose(dim0, dim1) → Tensor

See torch.transpose()

transpose_(dim0, dim1) → Tensor

In-place version of transpose()

triangular_solve(A, upper=True, transpose=False, unitriangular=False) -> (Tensor, Tensor)

See torch.triangular_solve()

tril(k=0) → Tensor

See torch.tril()

tril_(k=0) → Tensor

In-place version of tril()

triu(k=0) → Tensor

See torch.triu()

triu_(k=0) → Tensor

In-place version of triu()

trunc() → Tensor

See torch.trunc()

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())

Parameters

tensor (Tensor) – the tensor which has the desired type

unbind(dim=0) → seq

See torch.unbind()

unfold(dimension, size, step) → Tensor

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

Step between two slices is given by step.

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

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

Parameters
  • dimension (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)=1tofromP(x) = \dfrac{1}{\text{to} - \text{from}}
unique(sorted=True, return_inverse=False, return_counts=False, dim=None)[source]

Returns the unique elements of the input tensor.

See torch.unique()

unique_consecutive(return_inverse=False, return_counts=False, dim=None)[source]

Eliminates all but the first element from every consecutive group of equivalent elements.

See torch.unique_consecutive()

unsqueeze(dim) → Tensor

See torch.unsqueeze()

unsqueeze_(dim) → Tensor

In-place version of unsqueeze()

values() → Tensor

If self is a sparse COO tensor (i.e., with torch.sparse_coo layout), this returns a view of the contained values tensor. Otherwise, this throws an error.

See also Tensor.indices().

Note

This method can only be called on a coalesced sparse tensor. See Tensor.coalesce() for details.

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

See torch.var()

view(*shape) → Tensor

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

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,,d+kd, d+1, \dots, d+k that satisfy the following contiguity-like condition that i=0,,k1\forall i = 0, \dots, k-1 ,

stride[i]=stride[i+1]×size[i+1]\text{stride}[i] = \text{stride}[i+1] \times \text{size}[i+1]

Otherwise, contiguous() needs to be called before the tensor can be viewed. See also: reshape(), which returns a view if the shapes are compatible, and copies (equivalent to calling contiguous()) otherwise.

Parameters

shape (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])

>>> a = torch.randn(1, 2, 3, 4)
>>> a.size()
torch.Size([1, 2, 3, 4])
>>> b = a.transpose(1, 2)  # Swaps 2nd and 3rd dimension
>>> b.size()
torch.Size([1, 3, 2, 4])
>>> c = a.view(1, 3, 2, 4)  # Does not change tensor layout in memory
>>> c.size()
torch.Size([1, 3, 2, 4])
>>> torch.equal(b, c)
False
view_as(other) → Tensor

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

Please see view() for more information about view.

Parameters

other (torch.Tensor) – The result tensor has the same size as other.

where(condition, y) → Tensor

self.where(condition, y) is equivalent to torch.where(condition, self, y). See torch.where()

zero_() → Tensor

Fills self tensor with zeros.

class torch.BoolTensor

The following methods are unique to torch.BoolTensor.

all()
all() → bool

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

Example:

>>> a = torch.rand(1, 2).bool()
>>> a
tensor([[False, True]], dtype=torch.bool)
>>> a.all()
tensor(False, dtype=torch.bool)
all(dim, keepdim=False, out=None) → Tensor

Returns True if all elements in each row of the tensor in the given dimension dim are True, False otherwise.

If keepdim is True, the output tensor is of the same size as input except in the dimension dim where it is of size 1. Otherwise, dim is squeezed (see torch.squeeze()), resulting in the output tensor having 1 fewer dimension than input.

Parameters
  • dim (int) – the dimension to reduce

  • keepdim (bool) – whether the output tensor has dim retained or not

  • out (Tensor, optional) – the output tensor

Example:

>>> a = torch.rand(4, 2).bool()
>>> a
tensor([[True, True],
        [True, False],
        [True, True],
        [True, True]], dtype=torch.bool)
>>> a.all(dim=1)
tensor([ True, False,  True,  True], dtype=torch.bool)
>>> a.all(dim=0)
tensor([ True, False], dtype=torch.bool)
any()
any() → bool

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

Example:

>>> a = torch.rand(1, 2).bool()
>>> a
tensor([[False, True]], dtype=torch.bool)
>>> a.any()
tensor(True, dtype=torch.bool)
any(dim, keepdim=False, out=None) → Tensor

Returns True if any elements in each row of the tensor in the given dimension dim are True, False otherwise.

If keepdim is True, the output tensor is of the same size as input except in the dimension dim where it is of size 1. Otherwise, dim is squeezed (see torch.squeeze()), resulting in the output tensor having 1 fewer dimension than input.

Parameters
  • dim (int) – the dimension to reduce

  • keepdim (bool) – whether the output tensor has dim retained or not

  • out (Tensor, optional) – the output tensor

Example:

>>> a = torch.randn(4, 2) < 0
>>> a
tensor([[ True,  True],
        [False,  True],
        [ True,  True],
        [False, False]])
>>> a.any(1)
tensor([ True,  True,  True, False])
>>> a.any(0)
tensor([True, True])

Docs

Access comprehensive developer documentation for PyTorch

View Docs

Tutorials

Get in-depth tutorials for beginners and advanced developers

View Tutorials

Resources

Find development resources and get your questions answered

View Resources