Pytorch基础语法：张量tensor

TORCH.TENSOR

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

张量是一个包含单一数据类型元素的多维矩阵。

Torch定义了10种带有CPU和GPU变体的张量类型，如下所示:

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 1	torch.float16 or torch.half	torch.HalfTensor	torch.cuda.HalfTensor
16-bit floating point 2	torch.bfloat16	torch.BFloat16Tensor	torch.cuda.BFloat16Tensor
32-bit complex	torch.complex32
64-bit complex	torch.complex64
128-bit complex	torch.complex128 or torch.cdouble
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
quantized 8-bit integer (unsigned)	torch.quint8	torch.ByteTensor	/
quantized 8-bit integer (signed)	torch.qint8	torch.CharTensor	/
quantized 32-bit integer (signed)	torch.qfint32	torch.IntTensor	/
quantized 4-bit integer (unsigned) 3	torch.quint4x2	torch.ByteTensor	/

1

Sometimes referred to as binary16: uses 1 sign, 5 exponent, and 10 significand bits. Useful when precision is important at the expense of range.

有时也称为binary16:使用1个符号，5个指数和10个有效位。当以牺牲距离为代价求精度时很有用。

2

Sometimes referred to as Brain Floating Point: uses 1 sign, 8 exponent, and 7 significand bits. Useful when range is important, since it has the same number of exponent bits as float32

有时被称为脑浮点数:使用1个符号，8个指数，7个有效位:Four pieces当range很重要时很有用，因为它的指数位数与float32相同

3

quantized 4-bit integer is stored as a 8-bit signed integer. Currently it’s only supported in EmbeddingBag operator.

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

量化的4位整数存储为8位有符号整数。目前它只支持在embedded bag操作系统。torch.tensor是默认张量类型(torch.FloatTensor)。

初始化和基本操作

张量可以通过使用torch.张量()构造函数从Python列表或序列构造。

import torch
# print(torch.cuda.is_available())
print(torch.tensor([[1., -1.], [1., -1.]]))
print(torch.tensor(np.array([[1, 2, 3], [4, 5, 6]])))

运行结果：
D:\conda_env\python.exe D:/session9/test.py
tensor([[ 1., -1.],
Traceback (most recent call last):
        [ 1., -1.]])
  File "D:/session9/test.py", line 4, in 
    print(torch.tensor(np.array([[1, 2, 3], [4, 5, 6]])))
NameError: name 'np' is not defined

Process finished with exit code 1

出现错误：NameError: name 'np' is not defined

解决办法：使用import导入该模块进行使用，为了方便编写代码，在导入时同时给numpy一个别名。之后在使用该模块时，可以使用缩写的别名进行调用。

import numpy as np

注意：

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:

特定数据类型的张量可以通过传递一个torch.dtyper来构造或一个torch.device到构造函数或张量创建操作:

import torch
import numpy as np
# print(torch.cuda.is_available())
print(torch.tensor([[1., -1.], [1., -1.]]))
print(torch.tensor(np.array([[1, 2, 3], [4, 5, 6]])))
print(torch.zeros([2, 4], dtype=torch.int32))#定义一个两行四列的零矩阵
cuda0 = torch.device('cuda:0')
print(torch.ones([2, 4], dtype=torch.float64, device=cuda0))

运行结果：
D:\conda_env\python.exe D:/session9/test.py
tensor([[ 1., -1.],
        [ 1., -1.]])
tensor([[1, 2, 3],
        [4, 5, 6]], dtype=torch.int32)
tensor([[0, 0, 0, 0],
        [0, 0, 0, 0]], dtype=torch.int32)
tensor([[1., 1., 1., 1.],
        [1., 1., 1., 1.]], device='cuda:0', dtype=torch.float64)

Process finished with exit code 0

如果你想了解 torch.device(“cuda”) 与 torch.device(“cuda:0”)之间的不同，点击下面链接：

Difference between torch.device("cuda") and torch.device("cuda:0") - PyTorch Forums

有关构建张量的更多信息，请参见  Creation Ops

张量的内容可以使用Python的索引和切片表示法来访问和修改:

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

运行结果：
D:\conda_env\python.exe D:/session9/test.py
tensor(6)

Process finished with exit code 0

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

运行结果：
D:\conda_env\python.exe D:/session9/test.py
tensor(6)
tensor([[1, 8, 3],
        [4, 5, 6]])

Process finished with exit code 0

使用torch.Tensor.item()从一个包含单个值的张量中获取一个Python数字:

import torch
x = torch.tensor([[1]])
print(x)
print(x.item())

运行结果：
D:\conda_env\python.exe D:/session9/test.py
tensor([[1]])
1

Process finished with exit code 0

import torch
x = torch.tensor(2.5)
print(x)
print(x.item())

运行结果：
D:\conda_env\python.exe D:/session9/test.py
tensor(2.5000)
2.5

Process finished with exit code 0

有关索引的更多信息，请参阅有关索引的更多信息，请参阅Indexing, Slicing, Joining, Mutating Ops

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

一个张量可以用requires_grade - true创建，这样火炬。Autograd记录对它们的操作，以便自动区分。

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

运行结果：
D:\conda_env\python.exe D:/session9/test.py
tensor([[ 2., -4.],
        [ 2.,  2.]])

Process finished with exit code 0

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

每个张量都有一个相关的torch.Storage，用来保存数据。张量类还提供了多维的、跨越式的存储视图，并定义了对它的数值运算

备注：有关张量视图的更多信息，请参见 Tensor Views.

请注意了解更多关于torch.dtype, torch.device,和torch.layout。了解 torch.Tensor的属性参见 Tensor Attributes.

注意：改变张量的方法用下划线后缀标记。例如，torch.FloatTensor.abs()会计算che的绝对值并返回修改后的张量，而torch.FloatTensor.abs()会计算一个新张量的结果

说明：改变一个现有张量的torch.device and/or torch.dtype, 考虑在tenson上使用to()方法

警告：目前执行的torch.Tensor引入了内存开销，因此它可能会在有许多微小张量的应用程序中导致意想不到的高内存使用量。如果这是你的情况，考虑使用一个大型结构。

张量类引用

CLASS torch.Tensor

有几种主要的方法来创建一个张量，这取决于你的用例:

• 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.

• Tensor.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).

  Tensor.new_tensor Returns a new Tensor with data as the tensor data. Tensor.new_full Returns a Tensor of size size filled with fill_value. Tensor.new_empty Returns a Tensor of size size filled with uninitialized data. Tensor.new_ones Returns a Tensor of size size filled with 1. Tensor.new_zeros Returns a Tensor of size size filled with 0. Tensor.is_cuda Is True if the Tensor is stored on the GPU, False otherwise. Tensor.is_quantized Is True if the Tensor is quantized, False otherwise. Tensor.is_meta Is True if the Tensor is a meta tensor, False otherwise. Tensor.device Is the torch.device where this Tensor is. Tensor.grad This attribute is None by default and becomes a Tensor the first time a call to backward() computes gradients for self. Tensor.ndim Alias for dim() Tensor.real Returns a new tensor containing real values of the self tensor. Tensor.imag Returns a new tensor containing imaginary values of the self tensor. Tensor.abs See torch.abs() Tensor.abs_ In-place version of abs() Tensor.absolute Alias for abs() Tensor.absolute_ In-place version of absolute() Alias for abs_() Tensor.acos See torch.acos() Tensor.acos_ In-place version of acos() Tensor.arccos See torch.arccos() Tensor.arccos_ In-place version of arccos() Tensor.add Add a scalar or tensor to self tensor. Tensor.add_ In-place version of add() Tensor.addbmm See torch.addbmm() Tensor.addbmm_ In-place version of addbmm() Tensor.addcdiv See torch.addcdiv() Tensor.addcdiv_ In-place version of addcdiv() Tensor.addcmul See torch.addcmul() Tensor.addcmul_ In-place version of addcmul() Tensor.addmm See torch.addmm() Tensor.addmm_ In-place version of addmm() Tensor.sspaddmm See torch.sspaddmm() Tensor.addmv See torch.addmv() Tensor.addmv_ In-place version of addmv() Tensor.addr See torch.addr() Tensor.addr_ In-place version of addr() Tensor.allclose See torch.allclose() Tensor.amax See torch.amax() Tensor.amin See torch.amin() Tensor.aminmax See torch.aminmax() Tensor.angle See torch.angle() Tensor.apply_ Applies the function callable to each element in the tensor, replacing each element with the value returned by callable. Tensor.argmax See torch.argmax() Tensor.argmin See torch.argmin() Tensor.argsort See torch.argsort() Tensor.asin See torch.asin() Tensor.asin_ In-place version of asin() Tensor.arcsin See torch.arcsin() Tensor.arcsin_ In-place version of arcsin() Tensor.as_strided See torch.as_strided() Tensor.atan See torch.atan() Tensor.atan_ In-place version of atan() Tensor.arctan See torch.arctan() Tensor.arctan_ In-place version of arctan() Tensor.atan2 See torch.atan2() Tensor.atan2_ In-place version of atan2() Tensor.all See torch.all() Tensor.any See torch.any() Tensor.backward Computes the gradient of current tensor w.r.t. Tensor.baddbmm See torch.baddbmm() Tensor.baddbmm_ In-place version of baddbmm() Tensor.bernoulli Returns a result tensor where each \texttt{result[i]}result[i] is independently sampled from \text{Bernoulli}(\texttt{self[i]})Bernoulli(self[i]). Tensor.bernoulli_ Fills each location of self with an independent sample from \text{Bernoulli}(\texttt{p})Bernoulli(p). Tensor.bfloat16 self.bfloat16() is equivalent to self.to(torch.bfloat16). Tensor.bincount See torch.bincount() Tensor.bitwise_not See torch.bitwise_not() Tensor.bitwise_not_ In-place version of bitwise_not() Tensor.bitwise_and See torch.bitwise_and() Tensor.bitwise_and_ In-place version of bitwise_and() Tensor.bitwise_or See torch.bitwise_or() Tensor.bitwise_or_ In-place version of bitwise_or() Tensor.bitwise_xor See torch.bitwise_xor() Tensor.bitwise_xor_ In-place version of bitwise_xor() Tensor.bitwise_left_shift See torch.bitwise_left_shift() Tensor.bitwise_left_shift_ In-place version of bitwise_left_shift() Tensor.bitwise_right_shift See torch.bitwise_right_shift() Tensor.bitwise_right_shift_ In-place version of bitwise_right_shift() Tensor.bmm See torch.bmm() Tensor.bool self.bool() is equivalent to self.to(torch.bool). Tensor.byte self.byte() is equivalent to self.to(torch.uint8). Tensor.broadcast_to See torch.broadcast_to(). Tensor.cauchy_ Fills the tensor with numbers drawn from the Cauchy distribution: Tensor.ceil See torch.ceil() Tensor.ceil_ In-place version of ceil() Tensor.char self.char() is equivalent to self.to(torch.int8). Tensor.cholesky See torch.cholesky() Tensor.cholesky_inverse See torch.cholesky_inverse() Tensor.cholesky_solve See torch.cholesky_solve() Tensor.chunk See torch.chunk() Tensor.clamp See torch.clamp() Tensor.clamp_ In-place version of clamp() Tensor.clip Alias for clamp(). Tensor.clip_ Alias for clamp_(). Tensor.clone See torch.clone() Tensor.contiguous Returns a contiguous in memory tensor containing the same data as self tensor. Tensor.copy_ Copies the elements from src into self tensor and returns self. Tensor.conj See torch.conj() Tensor.conj_physical See torch.conj_physical() Tensor.conj_physical_ In-place version of conj_physical() Tensor.resolve_conj See torch.resolve_conj() Tensor.resolve_neg See torch.resolve_neg() Tensor.copysign See torch.copysign() Tensor.copysign_ In-place version of copysign() Tensor.cos See torch.cos() Tensor.cos_ In-place version of cos() Tensor.cosh See torch.cosh() Tensor.cosh_ In-place version of cosh() Tensor.corrcoef See torch.corrcoef() Tensor.count_nonzero See torch.count_nonzero() Tensor.cov See torch.cov() Tensor.acosh See torch.acosh() Tensor.acosh_ In-place version of acosh() Tensor.arccosh acosh() -> Tensor Tensor.arccosh_ acosh_() -> Tensor Tensor.cpu Returns a copy of this object in CPU memory. Tensor.cross See torch.cross() Tensor.cuda Returns a copy of this object in CUDA memory. Tensor.logcumsumexp See torch.logcumsumexp() Tensor.cummax See torch.cummax() Tensor.cummin See torch.cummin() Tensor.cumprod See torch.cumprod() Tensor.cumprod_ In-place version of cumprod() Tensor.cumsum See torch.cumsum() Tensor.cumsum_ In-place version of cumsum() Tensor.data_ptr Returns the address of the first element of self tensor. Tensor.deg2rad See torch.deg2rad() Tensor.dequantize Given a quantized Tensor, dequantize it and return the dequantized float Tensor. Tensor.det See torch.det() Tensor.dense_dim Return the number of dense dimensions in a sparse tensor self. Tensor.detach Returns a new Tensor, detached from the current graph. Tensor.detach_ Detaches the Tensor from the graph that created it, making it a leaf. Tensor.diag See torch.diag() Tensor.diag_embed See torch.diag_embed() Tensor.diagflat See torch.diagflat() Tensor.diagonal See torch.diagonal() Tensor.fill_diagonal_ Fill the main diagonal of a tensor that has at least 2-dimensions. Tensor.fmax See torch.fmax() Tensor.fmin See torch.fmin() Tensor.diff See torch.diff() Tensor.digamma See torch.digamma() Tensor.digamma_ In-place version of digamma() Tensor.dim Returns the number of dimensions of self tensor. Tensor.dist See torch.dist() Tensor.div See torch.div() Tensor.div_ In-place version of div() Tensor.divide See torch.divide() Tensor.divide_ In-place version of divide() Tensor.dot See torch.dot() Tensor.double self.double() is equivalent to self.to(torch.float64). Tensor.dsplit See torch.dsplit() Tensor.eig See torch.eig() Tensor.element_size Returns the size in bytes of an individual element. Tensor.eq See torch.eq() Tensor.eq_ In-place version of eq() Tensor.equal See torch.equal() Tensor.erf See torch.erf() Tensor.erf_ In-place version of erf() Tensor.erfc See torch.erfc() Tensor.erfc_ In-place version of erfc() Tensor.erfinv See torch.erfinv() Tensor.erfinv_ In-place version of erfinv() Tensor.exp See torch.exp() Tensor.exp_ In-place version of exp() Tensor.expm1 See torch.expm1() Tensor.expm1_ In-place version of expm1() Tensor.expand Returns a new view of the self tensor with singleton dimensions expanded to a larger size. Tensor.expand_as Expand this tensor to the same size as other. Tensor.exponential_ Fills self tensor with elements drawn from the exponential distribution: Tensor.fix See torch.fix(). Tensor.fix_ In-place version of fix() Tensor.fill_ Fills self tensor with the specified value. Tensor.flatten See torch.flatten() Tensor.flip See torch.flip() Tensor.fliplr See torch.fliplr() Tensor.flipud See torch.flipud() Tensor.float self.float() is equivalent to self.to(torch.float32). Tensor.float_power See torch.float_power() Tensor.float_power_ In-place version of float_power() Tensor.floor See torch.floor() Tensor.floor_ In-place version of floor() Tensor.floor_divide See torch.floor_divide() Tensor.floor_divide_ In-place version of floor_divide() Tensor.fmod See torch.fmod() Tensor.fmod_ In-place version of fmod() Tensor.frac See torch.frac() Tensor.frac_ In-place version of frac() Tensor.frexp See torch.frexp() Tensor.gather See torch.gather() Tensor.gcd See torch.gcd() Tensor.gcd_ In-place version of gcd() Tensor.ge See torch.ge(). Tensor.ge_ In-place version of ge(). Tensor.greater_equal See torch.greater_equal(). Tensor.greater_equal_ In-place version of greater_equal(). Tensor.geometric_ Fills self tensor with elements drawn from the geometric distribution: Tensor.geqrf See torch.geqrf() Tensor.ger See torch.ger() Tensor.get_device For CUDA tensors, this function returns the device ordinal of the GPU on which the tensor resides. Tensor.gt See torch.gt(). Tensor.gt_ In-place version of gt(). Tensor.greater See torch.greater(). Tensor.greater_ In-place version of greater(). Tensor.half self.half() is equivalent to self.to(torch.float16). Tensor.hardshrink See torch.nn.functional.hardshrink() Tensor.heaviside See torch.heaviside() Tensor.histc See torch.histc() Tensor.histogram See torch.histogram() Tensor.hsplit See torch.hsplit() Tensor.hypot See torch.hypot() Tensor.hypot_ In-place version of hypot() Tensor.i0 See torch.i0() Tensor.i0_ In-place version of i0() Tensor.igamma See torch.igamma() Tensor.igamma_ In-place version of igamma() Tensor.igammac See torch.igammac() Tensor.igammac_ In-place version of igammac() Tensor.index_add_ Accumulate the elements of alpha times tensor into the self tensor by adding to the indices in the order given in index. Tensor.index_add Out-of-place version of torch.Tensor.index_add_(). Tensor.index_copy_ Copies the elements of tensor into the self tensor by selecting the indices in the order given in index. Tensor.index_copy Out-of-place version of torch.Tensor.index_copy_(). Tensor.index_fill_ Fills the elements of the self tensor with value value by selecting the indices in the order given in index. Tensor.index_fill Out-of-place version of torch.Tensor.index_fill_(). Tensor.index_put_ Puts values from the tensor values into the tensor self using the indices specified in indices (which is a tuple of Tensors). Tensor.index_put Out-place version of index_put_(). Tensor.index_select See torch.index_select() Tensor.indices Return the indices tensor of a sparse COO tensor. Tensor.inner See torch.inner(). Tensor.int self.int() is equivalent to self.to(torch.int32). Tensor.int_repr 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. Tensor.inverse See torch.inverse() Tensor.isclose See torch.isclose() Tensor.isfinite See torch.isfinite() Tensor.isinf See torch.isinf() Tensor.isposinf See torch.isposinf() Tensor.isneginf See torch.isneginf() Tensor.isnan See torch.isnan() Tensor.is_contiguous Returns True if self tensor is contiguous in memory in the order specified by memory format. Tensor.is_complex Returns True if the data type of self is a complex data type. Tensor.is_conj Returns True if the conjugate bit of self is set to true. Tensor.is_floating_point Returns True if the data type of self is a floating point data type. Tensor.is_inference See torch.is_inference() Tensor.is_leaf All Tensors that have requires_grad which is False will be leaf Tensors by convention. Tensor.is_pinned Returns true if this tensor resides in pinned memory. Tensor.is_set_to Returns True if both tensors are pointing to the exact same memory (same storage, offset, size and stride). Tensor.is_shared Checks if tensor is in shared memory. Tensor.is_signed Returns True if the data type of self is a signed data type. Tensor.is_sparse Is True if the Tensor uses sparse storage layout, False otherwise. Tensor.istft See torch.istft() Tensor.isreal See torch.isreal() Tensor.item Returns the value of this tensor as a standard Python number. Tensor.kthvalue See torch.kthvalue() Tensor.lcm See torch.lcm() Tensor.lcm_ In-place version of lcm() Tensor.ldexp See torch.ldexp() Tensor.ldexp_ In-place version of ldexp() Tensor.le See torch.le(). Tensor.le_ In-place version of le(). Tensor.less_equal See torch.less_equal(). Tensor.less_equal_ In-place version of less_equal(). Tensor.lerp See torch.lerp() Tensor.lerp_ In-place version of lerp() Tensor.lgamma See torch.lgamma() Tensor.lgamma_ In-place version of lgamma() Tensor.log See torch.log() Tensor.log_ In-place version of log() Tensor.logdet See torch.logdet() Tensor.log10 See torch.log10() Tensor.log10_ In-place version of log10() Tensor.log1p See torch.log1p() Tensor.log1p_ In-place version of log1p() Tensor.log2 See torch.log2() Tensor.log2_ In-place version of log2() Tensor.log_normal_ Fills self tensor with numbers samples from the log-normal distribution parameterized by the given mean \muμ and standard deviation \sigmaσ. Tensor.logaddexp See torch.logaddexp() Tensor.logaddexp2 See torch.logaddexp2() Tensor.logsumexp See torch.logsumexp() Tensor.logical_and See torch.logical_and() Tensor.logical_and_ In-place version of logical_and() Tensor.logical_not See torch.logical_not() Tensor.logical_not_ In-place version of logical_not() Tensor.logical_or See torch.logical_or() Tensor.logical_or_ In-place version of logical_or() Tensor.logical_xor See torch.logical_xor() Tensor.logical_xor_ In-place version of logical_xor() Tensor.logit See torch.logit() Tensor.logit_ In-place version of logit() Tensor.long self.long() is equivalent to self.to(torch.int64). Tensor.lstsq See torch.lstsq() Tensor.lt See torch.lt(). Tensor.lt_ In-place version of lt(). Tensor.less lt(other) -> Tensor Tensor.less_ In-place version of less(). Tensor.lu See torch.lu() Tensor.lu_solve See torch.lu_solve() Tensor.as_subclass Makes a cls instance with the same data pointer as self. Tensor.map_ Applies callable for each element in self tensor and the given tensor and stores the results in self tensor. Tensor.masked_scatter_ Copies elements from source into self tensor at positions where the mask is True. Tensor.masked_scatter Out-of-place version of torch.Tensor.masked_scatter_() Tensor.masked_fill_ Fills elements of self tensor with value where mask is True. Tensor.masked_fill Out-of-place version of torch.Tensor.masked_fill_() Tensor.masked_select See torch.masked_select() Tensor.matmul See torch.matmul() Tensor.matrix_power NOTE matrix_power() is deprecated, use torch.linalg.matrix_power() instead. Tensor.matrix_exp See torch.matrix_exp() Tensor.max See torch.max() Tensor.maximum See torch.maximum() Tensor.mean See torch.mean() Tensor.nanmean See torch.nanmean() Tensor.median See torch.median() Tensor.nanmedian See torch.nanmedian() Tensor.min See torch.min() Tensor.minimum See torch.minimum() Tensor.mm See torch.mm() Tensor.smm See torch.smm() Tensor.mode See torch.mode() Tensor.movedim See torch.movedim() Tensor.moveaxis See torch.moveaxis() Tensor.msort See torch.msort() Tensor.mul See torch.mul(). Tensor.mul_ In-place version of mul(). Tensor.multiply See torch.multiply(). Tensor.multiply_ In-place version of multiply(). Tensor.multinomial See torch.multinomial() Tensor.mv See torch.mv() Tensor.mvlgamma See torch.mvlgamma() Tensor.mvlgamma_ In-place version of mvlgamma() Tensor.nansum See torch.nansum() Tensor.narrow See torch.narrow() Tensor.narrow_copy Same as Tensor.narrow() except returning a copy rather than shared storage. Tensor.ndimension Alias for dim() Tensor.nan_to_num See torch.nan_to_num(). Tensor.nan_to_num_ In-place version of nan_to_num(). Tensor.ne See torch.ne(). Tensor.ne_ In-place version of ne(). Tensor.not_equal See torch.not_equal(). Tensor.not_equal_ In-place version of not_equal(). Tensor.neg See torch.neg() Tensor.neg_ In-place version of neg() Tensor.negative See torch.negative() Tensor.negative_ In-place version of negative() Tensor.nelement Alias for numel() Tensor.nextafter See torch.nextafter() Tensor.nextafter_ In-place version of nextafter() Tensor.nonzero See torch.nonzero() Tensor.norm See torch.norm() Tensor.normal_ Fills self tensor with elements samples from the normal distribution parameterized by mean and std. Tensor.numel See torch.numel() Tensor.numpy Returns self tensor as a NumPy ndarray. Tensor.orgqr See torch.orgqr() Tensor.ormqr See torch.ormqr() Tensor.outer See torch.outer(). Tensor.permute See torch.permute() Tensor.pin_memory Copies the tensor to pinned memory, if it’s not already pinned. Tensor.pinverse See torch.pinverse() Tensor.polygamma See torch.polygamma() Tensor.polygamma_ In-place version of polygamma() Tensor.positive See torch.positive() Tensor.pow See torch.pow() Tensor.pow_ In-place version of pow() Tensor.prod See torch.prod() Tensor.put_ Copies the elements from source into the positions specified by index. Tensor.qr See torch.qr() Tensor.qscheme Returns the quantization scheme of a given QTensor. Tensor.quantile See torch.quantile() Tensor.nanquantile See torch.nanquantile() Tensor.q_scale Given a Tensor quantized by linear(affine) quantization, returns the scale of the underlying quantizer(). Tensor.q_zero_point Given a Tensor quantized by linear(affine) quantization, returns the zero_point of the underlying quantizer(). Tensor.q_per_channel_scales Given a Tensor quantized by linear (affine) per-channel quantization, returns a Tensor of scales of the underlying quantizer. Tensor.q_per_channel_zero_points Given a Tensor quantized by linear (affine) per-channel quantization, returns a tensor of zero_points of the underlying quantizer. Tensor.q_per_channel_axis Given a Tensor quantized by linear (affine) per-channel quantization, returns the index of dimension on which per-channel quantization is applied. Tensor.rad2deg See torch.rad2deg() Tensor.random_ Fills self tensor with numbers sampled from the discrete uniform distribution over [from, to - 1]. Tensor.ravel see torch.ravel() Tensor.reciprocal See torch.reciprocal() Tensor.reciprocal_ In-place version of reciprocal() Tensor.record_stream Ensures that the tensor memory is not reused for another tensor until all current work queued on stream are complete. Tensor.register_hook Registers a backward hook. Tensor.remainder See torch.remainder() Tensor.remainder_ In-place version of remainder() Tensor.renorm See torch.renorm() Tensor.renorm_ In-place version of renorm() Tensor.repeat Repeats this tensor along the specified dimensions. Tensor.repeat_interleave See torch.repeat_interleave(). Tensor.requires_grad Is True if gradients need to be computed for this Tensor, False otherwise. Tensor.requires_grad_ Change if autograd should record operations on this tensor: sets this tensor’s requires_grad attribute in-place. Tensor.reshape Returns a tensor with the same data and number of elements as self but with the specified shape. Tensor.reshape_as Returns this tensor as the same shape as other. Tensor.resize_ Resizes self tensor to the specified size. Tensor.resize_as_ Resizes the self tensor to be the same size as the specified tensor. Tensor.retain_grad Enables this Tensor to have their grad populated during backward(). Tensor.retains_grad Is True if this Tensor is non-leaf and its grad is enabled to be populated during backward(), False otherwise. Tensor.roll See torch.roll() Tensor.rot90 See torch.rot90() Tensor.round See torch.round() Tensor.round_ In-place version of round() Tensor.rsqrt See torch.rsqrt() Tensor.rsqrt_ In-place version of rsqrt() Tensor.scatter Out-of-place version of torch.Tensor.scatter_() Tensor.scatter_ Writes all values from the tensor src into self at the indices specified in the index tensor. Tensor.scatter_add_ Adds all values from the tensor other into self at the indices specified in the index tensor in a similar fashion as scatter_(). Tensor.scatter_add Out-of-place version of torch.Tensor.scatter_add_() Tensor.select Slices the self tensor along the selected dimension at the given index. Tensor.set_ Sets the underlying storage, size, and strides. Tensor.share_memory_ Moves the underlying storage to shared memory. Tensor.short self.short() is equivalent to self.to(torch.int16). Tensor.sigmoid See torch.sigmoid() Tensor.sigmoid_ In-place version of sigmoid() Tensor.sign See torch.sign() Tensor.sign_ In-place version of sign() Tensor.signbit See torch.signbit() Tensor.sgn See torch.sgn() Tensor.sgn_ In-place version of sgn() Tensor.sin See torch.sin() Tensor.sin_ In-place version of sin() Tensor.sinc See torch.sinc() Tensor.sinc_ In-place version of sinc() Tensor.sinh See torch.sinh() Tensor.sinh_ In-place version of sinh() Tensor.asinh See torch.asinh() Tensor.asinh_ In-place version of asinh() Tensor.arcsinh See torch.arcsinh() Tensor.arcsinh_ In-place version of arcsinh() Tensor.size Returns the size of the self tensor. Tensor.slogdet See torch.slogdet() Tensor.solve See torch.solve() Tensor.sort See torch.sort() Tensor.split See torch.split() Tensor.sparse_mask Returns a new sparse tensor with values from a strided tensor self filtered by the indices of the sparse tensor mask. Tensor.sparse_dim Return the number of sparse dimensions in a sparse tensor self. Tensor.sqrt See torch.sqrt() Tensor.sqrt_ In-place version of sqrt() Tensor.square See torch.square() Tensor.square_ In-place version of square() Tensor.squeeze See torch.squeeze() Tensor.squeeze_ In-place version of squeeze() Tensor.std See torch.std() Tensor.stft See torch.stft() Tensor.storage Returns the underlying storage. Tensor.storage_offset Returns self tensor’s offset in the underlying storage in terms of number of storage elements (not bytes). Tensor.storage_type Returns the type of the underlying storage. Tensor.stride Returns the stride of self tensor. Tensor.sub See torch.sub(). Tensor.sub_ In-place version of sub() Tensor.subtract See torch.subtract(). Tensor.subtract_ In-place version of subtract(). Tensor.sum See torch.sum() Tensor.sum_to_size Sum this tensor to size. Tensor.svd See torch.svd() Tensor.swapaxes See torch.swapaxes() Tensor.swapdims See torch.swapdims() Tensor.symeig See torch.symeig() Tensor.t See torch.t() Tensor.t_ In-place version of t() Tensor.tensor_split See torch.tensor_split() Tensor.tile See torch.tile() Tensor.to Performs Tensor dtype and/or device conversion. Tensor.to_mkldnn Returns a copy of the tensor in torch.mkldnn layout. Tensor.take See torch.take() Tensor.take_along_dim See torch.take_along_dim() Tensor.tan See torch.tan() Tensor.tan_ In-place version of tan() Tensor.tanh See torch.tanh() Tensor.tanh_ In-place version of tanh() Tensor.atanh See torch.atanh() Tensor.atanh_ In-place version of atanh() Tensor.arctanh See torch.arctanh() Tensor.arctanh_ In-place version of arctanh() Tensor.tolist Returns the tensor as a (nested) list. Tensor.topk See torch.topk() Tensor.to_sparse Returns a sparse copy of the tensor. Tensor.trace See torch.trace() Tensor.transpose See torch.transpose() Tensor.transpose_ In-place version of transpose() Tensor.triangular_solve See torch.triangular_solve() Tensor.tril See torch.tril() Tensor.tril_ In-place version of tril() Tensor.triu See torch.triu() Tensor.triu_ In-place version of triu() Tensor.true_divide See torch.true_divide() Tensor.true_divide_ In-place version of true_divide_() Tensor.trunc See torch.trunc() Tensor.trunc_ In-place version of trunc() Tensor.type Returns the type if dtype is not provided, else casts this object to the specified type. Tensor.type_as Returns this tensor cast to the type of the given tensor. Tensor.unbind See torch.unbind() Tensor.unfold Returns a view of the original tensor which contains all slices of size size from self tensor in the dimension dimension. Tensor.uniform_ Fills self tensor with numbers sampled from the continuous uniform distribution: Tensor.unique Returns the unique elements of the input tensor. Tensor.unique_consecutive Eliminates all but the first element from every consecutive group of equivalent elements. Tensor.unsqueeze See torch.unsqueeze() Tensor.unsqueeze_ In-place version of unsqueeze() Tensor.values Return the values tensor of a sparse COO tensor. Tensor.var See torch.var() Tensor.vdot See torch.vdot() Tensor.view Returns a new tensor with the same data as the self tensor but of a different shape. Tensor.view_as View this tensor as the same size as other. Tensor.vsplit See torch.vsplit() Tensor.where self.where(condition, y) is equivalent to torch.where(condition, self, y). Tensor.xlogy See torch.xlogy() Tensor.xlogy_ In-place version of xlogy() Tensor.zero_ Fills self tensor with zero