阅读(3427) (0)

PyTorch torch

2020-09-11 11:31:52 更新
译者:https//pytorch.org/docs/stable/torch.html

此外,它提供了许多实用程序,可有效地序列化张量和任意类型,以及其他有用的实用程序。

它具有CUDA对应项,使您能够具有计算能力> = 3.0的NVIDIA GPU上运行张量计算。

张量

torch.is_tensor(obj)¶

如果<cite> obj </ cite>是PyTorch张量,则返回True。

参数

obj(对象)–要测试的对象

torch.is_storage(obj)¶

如果<cite> obj </ cite>是PyTorch存储对象,则返回True。

参量

obj(对象)–要测试的对象

torch.is_floating_point(input) -> (bool)¶

如果input的数据类型是浮点数据类型,即torch.float64torch.float32torch.float16之一,则返回True。

参量

输入(tensor)–要测试的PyTorch张量

torch.set_default_dtype(d)¶

将默认浮点D型设置为d。类型该用作将  torch.tensor() 中类型推断的默认浮点类型。

默认浮点dtype最初为torch.float32

参量

d(torch.dtype)–浮点dtype,变为成为值

例:

>>> torch.tensor([1.2, 3]).dtype           # initial default for floating point is torch.float32
torch.float32
>>> torch.set_default_dtype(torch.float64)
>>> torch.tensor([1.2, 3]).dtype           # a new floating point tensor
torch.float64
torch.get_default_dtype() → torch.dtype¶

获取当前的默认浮点数  torch.dtype 。

例:

>>> torch.get_default_dtype()  # initial default for floating point is torch.float32
torch.float32
>>> torch.set_default_dtype(torch.float64)
>>> torch.get_default_dtype()  # default is now changed to torch.float64
torch.float64
>>> torch.set_default_tensor_type(torch.FloatTensor)  # setting tensor type also affects this
>>> torch.get_default_dtype()  # changed to torch.float32, the dtype for torch.FloatTensor
torch.float32
torch.set_default_tensor_type(t)¶

默认将的torch.Tensor类型设置为浮点张量类型t。类型该用作还将  torch.tensor() 中类型推断的默认浮点类型。

默认的浮点张量类型最初为torch.FloatTensor

参量

t(python:type或同轴)–浮点张量类型类型名称

例:

>>> torch.tensor([1.2, 3]).dtype    # initial default for floating point is torch.float32
torch.float32
>>> torch.set_default_tensor_type(torch.DoubleTensor)
>>> torch.tensor([1.2, 3]).dtype    # a new floating point tensor
torch.float64
torch.numel(input) → int¶

返回input张量中的元素总数。

参量

输入(张量)–输入张量。

例:

>>> a = torch.randn(1, 2, 3, 4, 5)
>>> torch.numel(a)
120
>>> a = torch.zeros(4,4)
>>> torch.numel(a)
16
torch.set_printoptions(precision=None, threshold=None, edgeitems=None, linewidth=None, profile=None, sci_mode=None)¶

设置打印选项。从NumPy无耻地拿走的物品

参量

  • precision –浮点输出的精度位数(默认= 4)。
  • 阈值 –触发汇总而不是完整的 <cite>repr</cite> 的数组元素总数(默认= 1000)。
  • edgeitems -每个维的开始和结束处摘要中数组项目的数量(默认= 3)。
  • 线宽 –用于插入换行符的每行字符数(默认= 80)。 阈值矩阵将忽略此参数。
  • 配置文件 – Sane 默认用于漂亮的打印。 可以使用以上任何选项覆盖。 (<cite>默认</cite>,<cite>短</cite>,<cite>完整</cite>中的任何一种)
  • sci_mode –启用(真)或禁用(假)科学计数法。 如果指定了 None(默认),则该值由 <cite>_Formatter</cite> 定义
torch.set_flush_denormal(mode) → bool¶

禁用 CPU 上的非正常浮​​点数。

如果您的系统支持刷新非正规数并且已成功配置刷新非正规模式,则返回True。 set_flush_denormal() 仅在支持 SSE3 的 x86 架构上受支持。

Parameters

模式 (bool )–控制是否启用冲洗非正常模式

Example:

>>> torch.set_flush_denormal(True)
True
>>> torch.tensor([1e-323], dtype=torch.float64)
tensor([ 0.], dtype=torch.float64)
>>> torch.set_flush_denormal(False)
True
>>> torch.tensor([1e-323], dtype=torch.float64)
tensor(9.88131e-324 *
       [ 1.0000], dtype=torch.float64)

创作行动

注意

随机抽样创建操作列在随机抽样下,包括: torch.rand() torch.rand_like()  torch.randn() torch.randn_like() torch.randint() torch.randint_like()  torch.randperm() 您也可以将 torch.empty() 与输入一起使用 位随机抽样方法来创建 torch.Tensor ,并从更广泛的分布范围内采样值。

torch.tensor(data, dtype=None, device=None, requires_grad=False, pin_memory=False) → Tensor¶

data构造一个张量。

警告

torch.tensor() 始终复制data。 如果您具有张量data并希望避免复制,请使用 torch.Tensor.requires_grad_() 或  torch.Tensor.detach() 。 如果您有 NumPy ndarray并想避免复制,请使用 torch.as_tensor() 。

Warning

当数据是张量 <cite>x</cite> 时, torch.tensor() 从传递的任何数据中读出“数据”,并构造一个叶子变量。 因此,torch.tensor(x)等同于x.clone().detach()torch.tensor(x, requires_grad=True)等同于x.clone().detach().requires_grad_(True)。 建议使用clone()detach()的等效项。

Parameters

  • 数据 (array_like )–张量的初始数据。 可以是列表,元组,NumPy ndarray,标量和其他类型。
  • dtype (torch.dtype ,可选)–返回张量的所需数据类型。 默认值:如果None,则从data推断数据类型。
  • 设备 (torch.device ,可选)–返回张量的所需设备。 默认值:如果None,则使用当前设备作为默认张量类型(请参见 torch.set_default_tensor_type())。 device将是用于 CPU 张量类型的 CPU,并且是用于 CUDA 张量类型的当前 CUDA 设备。
  • require_grad (bool , 可选)–如果 autograd 应该在返回的张量上记录操作。 默认值:False
  • pin_memory (bool , 可选)–如果设置,返回的张量将分配在固定的内存中。 仅适用于 CPU 张量。 默认值:False

Example:

>>> torch.tensor([[0.1, 1.2], [2.2, 3.1], [4.9, 5.2]])
tensor([[ 0.1000,  1.2000],
        [ 2.2000,  3.1000],
        [ 4.9000,  5.2000]])

>>> torch.tensor([0, 1])  # Type inference on data
tensor([ 0,  1])

>>> torch.tensor([[0.11111, 0.222222, 0.3333333]],
                 dtype=torch.float64,
                 device=torch.device('cuda:0'))  # creates a torch.cuda.DoubleTensor
tensor([[ 0.1111,  0.2222,  0.3333]], dtype=torch.float64, device='cuda:0')

>>> torch.tensor(3.14159)  # Create a scalar (zero-dimensional tensor)
tensor(3.1416)

>>> torch.tensor([])  # Create an empty tensor (of size (0,))
tensor([])
torch.sparse_coo_tensor(indices, values, size=None, dtype=None, device=None, requires_grad=False) → Tensor¶

在给定values和给定values的情况下,以非零元素构造 COO(rdinate)格式的稀疏张量。 稀疏张量可以是<cite>而不是</cite>,在那种情况下,索引中有重复的坐标,并且该索引处的值是所有重复值条目的总和: torch.sparse 。

Parameters

  • 索引 (array_like )–张量的初始数据。 可以是列表,元组,NumPy ndarray,标量和其他类型。 将在内部强制转换为torch.LongTensor。 索引是矩阵中非零值的坐标,因此应为二维,其中第一维是张量维数,第二维是非零值数。
  • 值 (array_like )–张量的初始值。 可以是列表,元组,NumPy ndarray,标量和其他类型。
  • 大小(列表,元组或torch.Size,可选)–稀疏张量的大小。 如果未提供,则将推断大小为足以容纳所有非零元素的最小大小。
  • dtype (torch.dtype ,可选)–返回张量的所需数据类型。默认值:如果为 None,则从values推断数据类型。
  • 设备 (torch.device ,可选)–返回张量的所需设备。 默认值:如果为 None,则使用当前设备作为默认张量类型(请参见 torch.set_default_tensor_type())。 device将是用于 CPU 张量类型的 CPU,是用于 CUDA 张量类型的当前 CUDA 设备。
  • requires_grad (bool__, optional) – If autograd should record operations on the returned tensor. Default: False.

Example:

>>> i = torch.tensor([[0, 1, 1],
                      [2, 0, 2]])
>>> v = torch.tensor([3, 4, 5], dtype=torch.float32)
>>> torch.sparse_coo_tensor(i, v, [2, 4])
tensor(indices=tensor([[0, 1, 1],
                       [2, 0, 2]]),
       values=tensor([3., 4., 5.]),
       size=(2, 4), nnz=3, layout=torch.sparse_coo)

>>> torch.sparse_coo_tensor(i, v)  # Shape inference
tensor(indices=tensor([[0, 1, 1],
                       [2, 0, 2]]),
       values=tensor([3., 4., 5.]),
       size=(2, 3), nnz=3, layout=torch.sparse_coo)

>>> torch.sparse_coo_tensor(i, v, [2, 4],
                            dtype=torch.float64,
                            device=torch.device('cuda:0'))
tensor(indices=tensor([[0, 1, 1],
                       [2, 0, 2]]),
       values=tensor([3., 4., 5.]),
       device='cuda:0', size=(2, 4), nnz=3, dtype=torch.float64,
       layout=torch.sparse_coo)

## Create an empty sparse tensor with the following invariants:
##   1\. sparse_dim + dense_dim = len(SparseTensor.shape)
##   2\. SparseTensor._indices().shape = (sparse_dim, nnz)
##   3\. SparseTensor._values().shape = (nnz, SparseTensor.shape[sparse_dim:])
#
## For instance, to create an empty sparse tensor with nnz = 0, dense_dim = 0 and
## sparse_dim = 1 (hence indices is a 2D tensor of shape = (1, 0))
>>> S = torch.sparse_coo_tensor(torch.empty([1, 0]), [], [1])
tensor(indices=tensor([], size=(1, 0)),
       values=tensor([], size=(0,)),
       size=(1,), nnz=0, layout=torch.sparse_coo)

## and to create an empty sparse tensor with nnz = 0, dense_dim = 1 and
## sparse_dim = 1
>>> S = torch.sparse_coo_tensor(torch.empty([1, 0]), torch.empty([0, 2]), [1, 2])
tensor(indices=tensor([], size=(1, 0)),
       values=tensor([], size=(0, 2)),
       size=(1, 2), nnz=0, layout=torch.sparse_coo)
torch.as_tensor(data, dtype=None, device=None) → Tensor¶

将数据转换为<cite>torch。张量</cite>。 如果数据已经是具有相同 <cite>dtype</cite> 和<cite>设备</cite>的<cite>张量</cite>,则不会执行任何复制,否则将使用新的<cite>张量</cite>。 如果数据<cite>张量</cite>>具有requires_grad=True,则返回保留计算图的计算图。 同样,如果数据是对应的 <cite>dtype</cite> 的ndarray,并且<cite>设备</cite>是 cpu,则不会执行任何复制。

Parameters

  • data (array_like) – Initial data for the tensor. Can be a list, tuple, NumPy ndarray,scalar, and other types.
  • dtype (torch.dtype, optional)  – the desired data type of returned tensor. Default:  if None, infers data type from data.
  • device(torch.device, optional) – the desired device of returned tensor. Default: if None, uses the current device for the default tensor type (see torch.set_default_tensor_type()). device will be the CPU for CPU tensor types and the current CUDA device for CUDA tensor types.

Example:

>>> a = numpy.array([1, 2, 3])
>>> t = torch.as_tensor(a)
>>> t
tensor([ 1,  2,  3])
>>> t[0] = -1
>>> a
array([-1,  2,  3])

>>> a = numpy.array([1, 2, 3])
>>> t = torch.as_tensor(a, device=torch.device('cuda'))
>>> t
tensor([ 1,  2,  3])
>>> t[0] = -1
>>> a
array([1,  2,  3])
torch.as_strided(input, size, stride, storage_offset=0) → Tensor¶

创建具有指定sizestridestorage_offset的现有<cite>炬管</cite> input的视图。

Warning

创建的张量中的一个以上元素可以引用单个存储位置。 结果,就地操作(尤其是矢量化的操作)可能会导致错误的行为。 如果需要写张量,请先克隆它们。

许多 PyTorch 函数可返回张量视图,并在此函数内部实现。 这些功能,例如 torch.Tensor.expand() ,更易于阅读,因此更可取。

Parameters

  • input (Tensor) – the input tensor.
  • 大小(元组 或 python:ints )–输出张量的形状
  • 跨度(元组 或 python:ints )–输出张量的跨度
  • storage_offset (python:int , 可选)–输出张量的基础存储中的偏移量

Example:

>>> x = torch.randn(3, 3)
>>> x
tensor([[ 0.9039,  0.6291,  1.0795],
        [ 0.1586,  2.1939, -0.4900],
        [-0.1909, -0.7503,  1.9355]])
>>> t = torch.as_strided(x, (2, 2), (1, 2))
>>> t
tensor([[0.9039, 1.0795],
        [0.6291, 0.1586]])
>>> t = torch.as_strided(x, (2, 2), (1, 2), 1)
tensor([[0.6291, 0.1586],
        [1.0795, 2.1939]])
torch.from_numpy(ndarray) → Tensor¶

numpy.ndarray创建 Tensor 。

返回的张量和ndarray共享相同的内存。 对张量的修改将反映在ndarray中,反之亦然。 返回的张量不可调整大小。

当前它接受具有numpy.float64numpy.float32numpy.float16numpy.int64numpy.int32numpy.int16numpy.int8numpy.uint8numpy.bool d 类型的ndarray

Example:

>>> a = numpy.array([1, 2, 3])
>>> t = torch.from_numpy(a)
>>> t
tensor([ 1,  2,  3])
>>> t[0] = -1
>>> a
array([-1,  2,  3])
torch.zeros(*size, out=None, dtype=None, layout=torch.strided, device=None, requires_grad=False) → Tensor¶

返回一个由标量值 <cite>0</cite> 填充的张量,其形状由变量参数size定义。

Parameters

  • 大小 (python:int ... )–定义输出张量形状的整数序列。 可以是可变数量的参数,也可以是列表或元组之类的集合。
  • 输出 (tensor , 可选)–输出张量。
  • dtype  (torch.dtype ,可选)–返回张量的所需数据类型。 默认值:如果None使用全局默认值(请参见 torch.set_default_tensor_type())。
  • 布局 (torch.layout ,可选)–返回的 Tensor 所需的布局。 默认值:torch.strided
  • device (torch.device, optional) – the desired device of returned tensor. Default: if None, uses the current device for the default tensor type (see torch.set_default_tensor_type()). device will be the CPU for CPU tensor types and the current CUDA device for CUDA tensor types.
  • requires_grad (bool__, optional) – If autograd should record operations on the returned tensor.Default: False.

Example:

>>> torch.zeros(2, 3)
tensor([[ 0.,  0.,  0.],
        [ 0.,  0.,  0.]])

>>> torch.zeros(5)
tensor([ 0.,  0.,  0.,  0.,  0.])
torch.zeros_like(input, dtype=None, layout=None, device=None, requires_grad=False) → Tensor¶

返回一个由标量值 <cite>0</cite> 填充的张量,其大小与input相同。 torch.zeros_like(input)等效于torch.zeros(input.size(), dtype=input.dtype, layout=input.layout, device=input.device)

Warning

从 0.4 开始,此功能不支持out关键字。 作为替代,旧的torch.zeros_like(input, out=output)等效于torch.zeros(input.size(), out=output)

Parameters

  • 输入 (tensor)– input的大小将确定输出张量的大小。
  • dtype (torch.dtype ,可选)–返回的 Tensor 的所需数据类型。 默认值:如果为None,则默认为input的dtype。
  • 布局 (torch.layout ,可选)–返回张量的所需布局。 默认值:如果为None,则默认为input的布局。
  • 设备 (torch.device ,可选)–返回张量的所需设备。 默认值:如果为None,则默认为input的设备。
  • requires_grad (bool__, optional) – If autograd should record operations on the returned tensor. Defaultt: False.

Example:

>>> input = torch.empty(2, 3)
>>> torch.zeros_like(input)
tensor([[ 0.,  0.,  0.],
        [ 0.,  0.,  0.]])
torch.ones(*size, out=None, dtype=None, layout=torch.strided, device=None, requires_grad=False) → Tensor¶

返回一个由标量值 <cite>1</cite> 填充的张量,其形状由变量自变量size定义。

Parameters

  • size (python:int...) – a sequence of integers defining the shape of the output tensor. Can be a variable number of arguments or a collection like a list or tuple.
  • out  (Tensor, optional) – the output tensor.
  • dtype (torch.dtype, optional) – the desired data type of returned tensor. Default: if None, uses a global default (see  torch.set_default_tensor_type() ).
  • layout  (torch.layout, optional) – the desired layout of returned Tensor. Default: torch.strided.
  • device(torch.device, optional) – the desired device of returned tensor. Default: if None, uses the current device for the default tensor type (see torch.set_default_tensor_type()). device will be the CPU for CPU tensor types and the current CUDA device for CUDA tensor types.
  • requires_grad (bool__, optional) – If autograd should record operations on the returned tensor. Default: False.

Example:

>>> torch.ones(2, 3)
tensor([[ 1.,  1.,  1.],
        [ 1.,  1.,  1.]])

>>> torch.ones(5)
tensor([ 1.,  1.,  1.,  1.,  1.])
torch.ones_like(input, dtype=None, layout=None, device=None, requires_grad=False) → Tensor¶

返回一个由标量值 <cite>1</cite> 填充的张量,其大小与input相同。 torch.ones_like(input)等效于torch.ones(input.size(), dtype=input.dtype, layout=input.layout, device=input.device)

Warning

从 0.4 开始,此功能不支持out关键字。 作为替代,旧的torch.ones_like(input, out=output)等效于torch.ones(input.size(), out=output)

Parameters

  • input (Tensor) – the size of input will determine size of the output tensor.
  • dtype(torch.dtype, optional)  – the desired data type of returned Tensor. Default: if None, defaults to the dtype of input.
  • layout(torch.layout, optional)– the desired layout of returned tensor. Default: if None, defaults to the layout of input.
  • device  (torch.device, optional)– the desired device of returned tensor. Default: if None, defaults to the device of input.
  • requires_grad (bool__, optional) – If autograd should record operations on the returned tensor.Default: False

Example:

>>> input = torch.empty(2, 3)
>>> torch.ones_like(input)
tensor([[ 1.,  1.,  1.],
        [ 1.,  1.,  1.]])
torch.arange(start=0, end, step=1, out=None, dtype=None, layout=torch.strided, device=None, requires_grad=False) → Tensor¶

返回大小为 的一维张量,该值具有从<cite>开始</cite>开始具有公共差step的间隔[start, end)的值。

请注意,与end比较时,非整数step会出现浮点舍入错误; 为了避免不一致,在这种情况下,建议在end中添加一个小的ε。

Parameters

  • 起始(编号)–点集的起始值。 默认值:0
  • 结束(编号)–点集的结束值
  • 步骤(编号)–每对相邻点之间的间隙。 默认值:1
  • out (Tensor, optional) – the output tensor.
  • dtype (torch.dtype ,可选)–返回张量的所需数据类型。 默认值:如果None使用全局默认值(请参阅 torch.set_default_tensor_type())。 如果未提供 <cite>dtype</cite> ,则从其他输入参数推断数据类型。 如果<cite>开始</cite>,<cite>结束</cite>或<cite>停止</cite>中的任何一个是浮点,则推断 <cite>dtype</cite> 为默认 dtype,请参见[ get_default_dtype() 。 否则,将 <cite>dtype</cite> 推断为 <cite>torch.int64</cite> 。
  • layout  (torch.layout, optional) – the desired layout of returned Tensor. Default: torch.strided.
  • device (torch.device, optional) – the desired device of returned tensor. Default: if None, uses the current device for the default tensor type (see torch.set_default_tensor_type()). device will be the CPU for CPU tensor types and the current CUDA device for CUDA tensor types.
  • requires_grad (bool__, optional) – If autograd should record operations on the returned tensor. Default False.

Example:

>>> torch.arange(5)
tensor([ 0,  1,  2,  3,  4])
>>> torch.arange(1, 4)
tensor([ 1,  2,  3])
>>> torch.arange(1, 2.5, 0.5)
tensor([ 1.0000,  1.5000,  2.0000])
torch.range(start=0, end, step=1, out=None, dtype=None, layout=torch.strided, device=None, requires_grad=False) → Tensor¶

在步骤step中返回大小为 的一维张量,其值从startend。 阶跃是张量中两个值之间的差距。

Warning

不推荐使用此功能,而推荐使用 torch.arange() 。

Parameters

  • start (python:float )–点集的起始值。 默认值:0
  • end (python:float )–点集的结束值
  • 步骤 (python:float )–每对相邻点之间的间隙。 默认值:1
  • out (Tensor, optional) – the output tensor.
  • dtype(torch.dtype, optional) – the desired data type of returned tensor. Default: if None, uses a global default (see  torch.set_default_tensor_type() ). If <cite>dtype</cite> is not given, infer the data type from the other input arguments. If any of <cite>start</cite>, <cite>end</cite>, or <cite>stop</cite> are floating-point, the <cite>dtype</cite> is inferred to be the default dtype, see get_default_dtype(). Otherwise, the <cite>dtype</cite> is inferred to be <cite>torch.int64</cite>.
  • layout  (torch.layout, optional) – the desired layout of returned Tensor. Default: torch.strided.
  • device  (torch.device, optional) – the desired device of returned tensor. Default:  if None, uses the current device for the default tensor type (see torch.set_default_tensor_type()). device will be the CPU for CPU tensor types and the current CUDA device for CUDA tensor types.
  • requires_grad (bool__, optional) – If autograd should record operations on the returned tensor. Default: False.

Example:

>>> torch.range(1, 4)
tensor([ 1.,  2.,  3.,  4.])
>>> torch.range(1, 4, 0.5)
tensor([ 1.0000,  1.5000,  2.0000,  2.5000,  3.0000,  3.5000,  4.0000])
torch.linspace(start, end, steps=100, out=None, dtype=None, layout=torch.strided, device=None, requires_grad=False) → Tensor¶

返回startend之间等距点的steps的一维张量。

输出张量为steps大小的 1-D。

Parameters

  • 开始 (python:float )–点集的起始值
  • end (python:float) – the ending value for the set of points
  • 步骤 (python:int )–在startend之间采样的点数。 默认值:100。
  • out (Tensor, optional) – the output tensor.
  • dtype (torch.dtype, optional) – the desired data type of returned tensor. Default: if None, uses a global default (see  torch.set_default_tensor_type() ).
  • layout  (torch.layout, optional) – the desired layout of returned Tensor. Default: torch.strided.
  • device (torch.device, optional) – the desired device of returned tensor. Default: if None, uses the current device for the default tensor type (see torch.set_default_tensor_type()). device will be the CPU for CPU tensor types and the current CUDA device for CUDA tensor types.
  • requires_grad (bool__, optional) – If autograd should record operations on the returned tensor. Default: False.

Example:

>>> torch.linspace(3, 10, steps=5)
tensor([  3.0000,   4.7500,   6.5000,   8.2500,  10.0000])
>>> torch.linspace(-10, 10, steps=5)
tensor([-10.,  -5.,   0.,   5.,  10.])
>>> torch.linspace(start=-10, end=10, steps=5)
tensor([-10.,  -5.,   0.,   5.,  10.])
>>> torch.linspace(start=-10, end=10, steps=1)
tensor([-10.])
torch.logspace(start, end, steps=100, base=10.0, out=None, dtype=None, layout=torch.strided, device=None, requires_grad=False) → Tensor¶

返回与 之间的底数base对数间隔的steps点的一维张量。

The output tensor is 1-D of size steps.

Parameters

  • start (python:float) – the starting value for the set of points
  • end (python:float) – the ending value for the set of points
  • steps (python:int) – number of points to sample between start and end. Default: 100.
  • 基数 (python:float )–对数函数的基数。 默认值:10.0
  • out (Tensor, optional) – the output tensor.
  • dtype (torch.dtype, optional) – the desired data type of returned tensor. Default: if None, uses a global default (see  torch.set_default_tensor_type() ).
  • layout (torch.layout, optional) – the desired layout of returned Tensor. Default: torch.strided.
  • device (torch.device, optional) – the desired device of returned tensor. Default: if None, uses the current device for the default tensor type (see torch.set_default_tensor_type()). device will be the CPU for CPU tensor types and the current CUDA device for CUDA tensor types.
  • requires_grad (bool__, optional) – If autograd should record operations on the returned tensor. Default: False.

Example:

>>> torch.logspace(start=-10, end=10, steps=5)
tensor([ 1.0000e-10,  1.0000e-05,  1.0000e+00,  1.0000e+05,  1.0000e+10])
>>> torch.logspace(start=0.1, end=1.0, steps=5)
tensor([  1.2589,   2.1135,   3.5481,   5.9566,  10.0000])
>>> torch.logspace(start=0.1, end=1.0, steps=1)
tensor([1.2589])
>>> torch.logspace(start=2, end=2, steps=1, base=2)
tensor([4.0])
torch.eye(n, m=None, out=None, dtype=None, layout=torch.strided, device=None, requires_grad=False) → Tensor¶

返回一个二维张量,对角线上有一个,其他位置为零。

Parameters

  • n (python:int )–行数
  • m (python:int , 可选)–默认为n的列数
  • out (Tensor, optional) – the output tensor.
  • dtype  (torch.dtype, optional) – the desired data type of returned tensor. Default: if None, uses a global default (see  torch.set_default_tensor_type() ).
  • layout  (torch.layout, optional) – the desired layout of returned Tensor. Default: torch.strided.
  • device (torch.device, optional) – the desired device of returned tensor. Default: if None, uses the current device for the default tensor type (see torch.set_default_tensor_type()). device will be the CPU for CPU tensor types and the current CUDA device for CUDA tensor types.
  • requires_grad (bool__, optional) – If autograd should record operations on the returned tensor. Default: False.

退货

二维张量,对角线上有一个,其他位置为零

返回类型

张量

Example:

>>> torch.eye(3)
tensor([[ 1.,  0.,  0.],
        [ 0.,  1.,  0.],
        [ 0.,  0.,  1.]])
torch.empty(*size, out=None, dtype=None, layout=torch.strided, device=None, requires_grad=False, pin_memory=False) → Tensor¶

返回填充有未初始化数据的张量。 张量的形状由变量参数size定义。

Parameters

  • size (python:int...) – a sequence of integers defining the shape of the output tensor. Can be a variable number of arguments or a collection like a list or tuple.
  • out (Tensor, optional) – the output tensor.
  • dtype (torch.dtype, optional) – the desired data type of returned tensor. Default: if None, uses a global default (see torch.set_default_tensor_type()).
  • layout (torch.layout, optional) – the desired layout of returned Tensor. Default: torch.strided.
  • device (torch.device, optional) – the desired device of returned tensor. Default: if None, uses the current device for the default tensor type (see torch.set_default_tensor_type()). device will be the CPU for CPU tensor types and the current CUDA device for CUDA tensor types.
  • requires_grad (bool__, optional) – If autograd should record operations on the returned tensor. Default: False.
  • pin_memory (bool__, optional) – If set, returned tensor would be allocated in the pinned memory. Works only for CPU tensors. Default: False.

Example:

>>> torch.empty(2, 3)
tensor(1.00000e-08 *
       [[ 6.3984,  0.0000,  0.0000],
        [ 0.0000,  0.0000,  0.0000]])
torch.empty_like(input, dtype=None, layout=None, device=None, requires_grad=False) → Tensor¶

返回与input相同大小的未初始化张量。 torch.empty_like(input)等效于torch.empty(input.size(), dtype=input.dtype, layout=input.layout, device=input.device)

Parameters

  • input (Tensor) – the size  of input will determine size of the output tensor.
  • dtype (torch.dtype, optional) – the desired data type of returned Tensor. Default: if None, defaults to the dtype of input.
  • layout  (torch.layout, optional) – the desired layout of returned tensor. Default: if None, defaults to the layout of input.
  • device(torch.device, optional) – the desired device of returned tensor. Default: if None, defaults to the device of input.
  • requires_grad (bool__, optional) – If autograd should record operations on the returned Default: False.

Example:

>>> torch.empty((2,3), dtype=torch.int64)
tensor([[ 9.4064e+13,  2.8000e+01,  9.3493e+13],
        [ 7.5751e+18,  7.1428e+18,  7.5955e+18]])
torch.empty_strided(size, stride, dtype=None, layout=None, device=None, requires_grad=False, pin_memory=False) → Tensor¶

返回填充有未初始化数据的张量。 张量的形状和步幅分别由变量参数sizestride定义。 torch.empty_strided(size, stride)等同于torch.empty(size).as_strided(size, stride)

Warning

创建的张量中的一个以上元素可以引用单个存储位置。 结果,就地操作(尤其是矢量化的操作)可能会导致错误的行为。 如果需要写张量,请先克隆它们。

Parameters

  • 大小(python:ints 的元组)–输出张量的形状
  • 跨度(python:ints 的元组)–输出张量的跨度
  • dtype (torch.dtype, optional) – the desired data type of returned tensor. Default: if None, uses a global default (see  torch.set_default_tensor_type() ).
  • layout  (torch.layout, optional) – the desired layout of returned Tensor. Default: torch.strided.
  • device (torch.device, optional) – the desired device of returned tensor. Default: if None, uses the current device for the default tensor type (see torch.set_default_tensor_type()). device will be the CPU for CPU tensor types and the current CUDA device for CUDA tensor types.
  • requires_grad (bool__, optional) – If autograd should record operations on the returned tensor. Default: False.
  • pin_memory (bool__, optional) – If set, returned tensor would be allocated in the pinned memory. Works only for CPU tensors. Default: False.

Example:

>>> a = torch.empty_strided((2, 3), (1, 2))
>>> a
tensor([[8.9683e-44, 4.4842e-44, 5.1239e+07],
        [0.0000e+00, 0.0000e+00, 3.0705e-41]])
>>> a.stride()
(1, 2)
>>> a.size()
torch.Size([2, 3])
torch.full(size, fill_value, out=None, dtype=None, layout=torch.strided, device=None, requires_grad=False) → Tensor¶

返回大小为size的张量,其中填充了fill_value

Parameters

  • 大小 (python:int ... )–定义输出张量形状的整数列表,元组或torch.Size
  • fill_value –用来填充输出张量的数字。
  • out (Tenso, optional) – the output tensor.
  • dtype (torch.dtype, optional) – the desired data type of returned tensor. Default: if None, uses a global default (see torch.set_default_tensor_type()).
  • layout (torch.layout, optional) – the desired layout of returned Tensor. Default: torch.strided.
  • device (torch.device, optional) – the desired device of returned tensor. Default: if None, uses the current device for the default tensor type (see torch.set_default_tensor_type()). device will be the CPU for CPU tensor types and the current CUDA device for CUDA tensor types.
  • requires_grad (bool__, optional) – If autograd should record operations on the returned tensor. Default: False.

Example:

>>> torch.full((2, 3), 3.141592)
tensor([[ 3.1416,  3.1416,  3.1416],
        [ 3.1416,  3.1416,  3.1416]])
torch.full_like(input, fill_value, out=None, dtype=None, layout=torch.strided, device=None, requires_grad=False) → Tensor¶

返回与填充有fill_valueinput大小相同的张量。 torch.full_like(input, fill_value)等同于torch.full(input.size(), fill_value, dtype=input.dtype, layout=input.layout, device=input.device)

Parameters

  • input (Tensor) – the size of input will determine size of the output tensor.
  • fill_value – the number to fill the output tensor with.
  • dtype  (torch.dtype, optional) – the desired data type of returned Tensor. Default: if None, defaults to the dtype of input.
  • layout (torch.layout, optional) – the desired layout of returned tensor. Default: if None, defaults to the layout of input.
  • device (torch.device, optional) – the desired device of returned tensor. Default: if None, defaults to the device of input.
  • requires_grad(bool__, optional) – If autograd should record operations on the returned tensor. Default: False.
torch.quantize_per_tensor(input, scale, zero_point, dtype) → Tensor¶

将浮点张量转换为具有给定比例和零点的量化张量。

Parameters

  • 输入 (tensor)–浮点张量进行量化
  • 标度 (python:float )–适用于量化公式的标度
  • zero_point (python:int )–映射为浮点零的整数值偏移
  • dtype (torch.dtype)–返回张量的所需数据类型。 必须是量化的 dtypes 之一:torch.quint8torch.qint8torch.qint32

Returns

新量化的张量

Return type

Tensor

Example:

>>> torch.quantize_per_tensor(torch.tensor([-1.0, 0.0, 1.0, 2.0]), 0.1, 10, torch.quint8)
tensor([-1.,  0.,  1.,  2.], size=(4,), dtype=torch.quint8,
       quantization_scheme=torch.per_tensor_affine, scale=0.1, zero_point=10)
>>> torch.quantize_per_tensor(torch.tensor([-1.0, 0.0, 1.0, 2.0]), 0.1, 10, torch.quint8).int_repr()
tensor([ 0, 10, 20, 30], dtype=torch.uint8)
torch.quantize_per_channel(input, scales, zero_points, axis, dtype) → Tensor¶

将浮点张量转换为具有给定比例和零点的每通道量化张量。

Parameters

  • input (Tensor) – float tensor to quantize
  • 秤 (tensor)–要使用的一维浮标秤,尺寸应匹配input.size(axis)
  • zero_points (python:int )–要使用的整数 1D 张量偏移量,大小应与input.size(axis)相匹配
  • 轴 (python:int )–应用每个通道量化的维度
  • dtype (torch.dtype) – the desired data type of returned tensor. Has to be one of the quantized dtypes: torch.quint8torch.qint8torch.qint32

Returns

A newly quantized tensor

Return type

Tensor

Example:

>>> x = torch.tensor([[-1.0, 0.0], [1.0, 2.0]])
>>> torch.quantize_per_channel(x, torch.tensor([0.1, 0.01]), torch.tensor([10, 0]), 0, torch.quint8)
tensor([[-1.,  0.],
        [ 1.,  2.]], size=(2, 2), dtype=torch.quint8,
       quantization_scheme=torch.per_channel_affine,
       scale=tensor([0.1000, 0.0100], dtype=torch.float64),
       zero_point=tensor([10,  0]), axis=0)
>>> torch.quantize_per_channel(x, torch.tensor([0.1, 0.01]), torch.tensor([10, 0]), 0, torch.quint8).int_repr()
tensor([[  0,  10],
        [100, 200]], dtype=torch.uint8)

索引,切片,联接,操作变更

torch.cat(tensors, dim=0, out=None) → Tensor¶

在给定维度上连接seq张量的给定序列。 所有张量必须具有相同的形状(在连接维中除外)或为空。

torch.cat() 可以看作是 torch.split() 和  torch.chunk()  的逆运算。

通过示例可以更好地理解 torch.cat() 。

Parameters

  • 张量(张量序列)–同一类型的任何 python 张量序列。 提供的非空张量必须具有相同的形状,但猫的尺寸除外。
  • 暗淡的 (python:int , 可选)–张量级联的尺寸
  • out (Tensor, optional) – the output tensor.

Example:

>>> x = torch.randn(2, 3)
>>> x
tensor([[ 0.6580, -1.0969, -0.4614],
        [-0.1034, -0.5790,  0.1497]])
>>> torch.cat((x, x, x), 0)
tensor([[ 0.6580, -1.0969, -0.4614],
        [-0.1034, -0.5790,  0.1497],
        [ 0.6580, -1.0969, -0.4614],
        [-0.1034, -0.5790,  0.1497],
        [ 0.6580, -1.0969, -0.4614],
        [-0.1034, -0.5790,  0.1497]])
>>> torch.cat((x, x, x), 1)
tensor([[ 0.6580, -1.0969, -0.4614,  0.6580, -1.0969, -0.4614,  0.6580,
         -1.0969, -0.4614],
        [-0.1034, -0.5790,  0.1497, -0.1034, -0.5790,  0.1497, -0.1034,
         -0.5790,  0.1497]])
torch.chunk(input, chunks, dim=0) → List of Tensors¶

将张量拆分为特定数量的块。

如果沿给定维度dim的张量大小不能被chunks整除,则最后一块将较小。

Parameters

  • 输入 (tensor)–要分割的张量
  • 块 (python:int )–要返回的块数
  • 暗淡的 (python:int )–沿其张量分裂的尺寸
torch.gather(input, dim, index, out=None, sparse_grad=False) → Tensor¶

沿<cite>昏暗</cite>指定的轴收集值。

对于 3-D 张量,输出指定为:

out[i][j][k] = input[index[i][j][k]][j][k]  # if dim == 0
out[i][j][k] = input[i][index[i][j][k]][k]  # if dim == 1
out[i][j][k] = input[i][j][index[i][j][k]]  # if dim == 2

如果input是大小为 a89ebd4ba33f6e8d4b992302696fcb6bdim = i的 n 维张量,则index必须是大小为  0998ac39fb5e46e656f2261d0af181b6的  5f0051b0454fa75ca446b59b47eff6f6-维张量,其中  5890f3c556cb9391345ace3ce2c49ff9out具有相同的大小 大小为index

Parameters

  • 输入 (tensor)–源张量
  • 暗淡的 (python:int )–沿其索引的轴
  • 索引 (LongTensor )–要收集的元素的索引
  • 输出 (tensor , 可选)–目标张量
  • sparse_grad (bool , 可选)–如果True,则梯度 w.r.t. input将是一个稀疏张量。

Example:

>>> t = torch.tensor([[1,2],[3,4]])
>>> torch.gather(t, 1, torch.tensor([[0,0],[1,0]]))
tensor([[ 1,  1],
        [ 4,  3]])
torch.index_select(input, dim, index, out=None) → Tensor¶

返回一个新张量,该张量使用index LongTensor 中的index中的条目沿维度dim索引input张量。

返回的张量具有与原始张量(input)相同的维数。 dim的尺寸与index的长度相同; 其他尺寸与原始张量中的尺寸相同。

Note

返回的张量不与原始张量使用相同的存储空间而不是。 如果out的形状与预期的形状不同,我们将默默地将其更改为正确的形状,并在必要时重新分配基础存储。

Parameters

  • input (Tensor) – the input tensor.
  • 暗淡的 (python:int )–我们索引的维度
  • 索引 (LongTensor )–包含要索引的索引的一维张量
  • out (Tensor, optional) – the output tensor.

Example:

>>> x = torch.randn(3, 4)
>>> x
tensor([[ 0.1427,  0.0231, -0.5414, -1.0009],
        [-0.4664,  0.2647, -0.1228, -1.1068],
        [-1.1734, -0.6571,  0.7230, -0.6004]])
>>> indices = torch.tensor([0, 2])
>>> torch.index_select(x, 0, indices)
tensor([[ 0.1427,  0.0231, -0.5414, -1.0009],
        [-1.1734, -0.6571,  0.7230, -0.6004]])
>>> torch.index_select(x, 1, indices)
tensor([[ 0.1427, -0.5414],
        [-0.4664, -0.1228],
        [-1.1734,  0.7230]])
torch.masked_select(input, mask, out=None) → Tensor¶

返回一个新的一维张量,该张量根据布尔值掩码mask为其 <cite>BoolTensor</cite> 索引input张量。

mask张量和input张量的形状不需要匹配,但它们必须是>可广播的。

Note

返回的张量是否而不是使用与原始张量相同的存储

Parameters

  • input (Tensor) – the input tensor.
  • 掩码 (ByteTensor )–包含二进制掩码的张量,以使用
  • out (Tensor, optional) – the output tensor.

Example:

>>> x = torch.randn(3, 4)
>>> x
tensor([[ 0.3552, -2.3825, -0.8297,  0.3477],
        [-1.2035,  1.2252,  0.5002,  0.6248],
        [ 0.1307, -2.0608,  0.1244,  2.0139]])
>>> mask = x.ge(0.5)
>>> mask
tensor([[False, False, False, False],
        [False, True, True, True],
        [False, False, False, True]])
>>> torch.masked_select(x, mask)
tensor([ 1.2252,  0.5002,  0.6248,  2.0139])
torch.narrow(input, dim, start, length) → Tensor¶

返回一个新的张量,该张量是input张量的缩小版本。 尺寸dimstart输入到start + length。 返回的张量和input张量共享相同的基础存储。

Parameters

  • 输入 (tensor)–张量变窄
  • 暗淡的 (python:int )–缩小范围
  • 开始 (python:int )–起始尺寸
  • 长度 (python:int )–到最终尺寸的距离

Example:

>>> x = torch.tensor([[1, 2, 3], [4, 5, 6], [7, 8, 9]])
>>> torch.narrow(x, 0, 0, 2)
tensor([[ 1,  2,  3],
        [ 4,  5,  6]])
>>> torch.narrow(x, 1, 1, 2)
tensor([[ 2,  3],
        [ 5,  6],
        [ 8,  9]])
torch.nonzero(input, *, out=None, as_tuple=False) → LongTensor or tuple of LongTensors¶

Note

torch.nonzero(..., as_tuple=False) (默认值)返回一个二维张量,其中每一行都是非零值的索引。

torch.nonzero(..., as_tuple=True) 返回一维索引张量的元组,允许进行高级索引,因此x[x.nonzero(as_tuple=True)]给出张量x的所有非零值。 在返回的元组中,每个索引张量都包含特定维度的非零索引。

有关这两种行为的更多详细信息,请参见下文。

 as_tuple “ False”(默认)时:

返回一个张量,该张量包含input所有非零元素的索引。 结果中的每一行都包含input中非零元素的索引。 结果按字典顺序排序,最后一个索引更改最快(C 样式)。

如果input具有 5f0051b0454fa75ca446b59b47eff6f6尺寸,则所得索引张量out的大小为 b6080a6ed8a7dd471ec6fd4b1023bc23,其中 e1bf4c09825257a3ffbe6bddb254bcb6input张量中非零元素的总数。

 as_tuple “ True” 时:

返回一维张量的元组,在input中每个维度一个张量,每个张量包含input所有非零元素的索引(在该维度中)。

如果input具有 5f0051b0454fa75ca446b59b47eff6f6尺寸,则生成的元组包含 e1bf4c09825257a3ffbe6bddb254bcb6大小的  5f0051b0454fa75ca446b59b47eff6f6张量,其中  e1bf4c09825257a3ffbe6bddb254bcb6input张量中非零元素的总数。

作为一种特殊情况,当input具有零维和非零标量值时,会将其视为具有一个元素的一维张量。

Parameters

  • input (Tensor) – the input tensor.
  • out (LongTensor , 可选)–包含索引的输出张量

Returns

如果as_tupleFalse,则包含索引的输出张量。 如果as_tupleTrue,则每个维度都有一个 1-D 张量,其中包含沿着该维度的每个非零元素的索引。

Return type

LongTensor 或 LongTensor 的元组

Example:

>>> torch.nonzero(torch.tensor([1, 1, 1, 0, 1]))
tensor([[ 0],
        [ 1],
        [ 2],
        [ 4]])
>>> torch.nonzero(torch.tensor([[0.6, 0.0, 0.0, 0.0],
                                [0.0, 0.4, 0.0, 0.0],
                                [0.0, 0.0, 1.2, 0.0],
                                [0.0, 0.0, 0.0,-0.4]]))
tensor([[ 0,  0],
        [ 1,  1],
        [ 2,  2],
        [ 3,  3]])
>>> torch.nonzero(torch.tensor([1, 1, 1, 0, 1]), as_tuple=True)
(tensor([0, 1, 2, 4]),)
>>> torch.nonzero(torch.tensor([[0.6, 0.0, 0.0, 0.0],
                                [0.0, 0.4, 0.0, 0.0],
                                [0.0, 0.0, 1.2, 0.0],
                                [0.0, 0.0, 0.0,-0.4]]), as_tuple=True)
(tensor([0, 1, 2, 3]), tensor([0, 1, 2, 3]))
>>> torch.nonzero(torch.tensor(5), as_tuple=True)
(tensor([0]),)
torch.reshape(input, shape) → Tensor¶

返回具有与input相同的数据和元素数量,但具有指定形状的张量。 如果可能,返回的张量将是input的视图。 否则,它将是副本。 连续输入和具有兼容步幅的输入可以在不复制的情况下进行重塑,但是您不应该依赖复制与查看行为。

当可以返回视图时,请参见 torch.Tensor.view() 。

单个尺寸可能为-1,在这种情况下,它是根据input中的其余尺寸和元素数量推断出来的。

Parameters

  • 输入 (tensor)–要重塑的张量
  • 形状 (python:ints 的元组)–新形状

Example:

>>> a = torch.arange(4.)
>>> torch.reshape(a, (2, 2))
tensor([[ 0.,  1.],
        [ 2.,  3.]])
>>> b = torch.tensor([[0, 1], [2, 3]])
>>> torch.reshape(b, (-1,))
tensor([ 0,  1,  2,  3])
torch.split(tensor, split_size_or_sections, dim=0)¶

将张量拆分为多个块。

如果split_size_or_sections是整数类型,则 tensor 将被拆分为大小相等的块(如果可能)。 如果沿给定维度dim的张量大小不能被split_size整除,则最后一个块将较小。

如果split_size_or_sections是列表,则根据split_size_or_sections将 tensor 拆分为dim,大小为dim

Parameters

  • 张量 (tensor)–张量分裂。
  • split_size_or_sections (python:int )或 ( 列表 ( python :int ))–单个块的大小或每个块的大小列表
  • 暗淡的 (python:int )–沿其张量分裂的尺寸。
torch.squeeze(input, dim=None, out=None) → Tensor¶

返回一个张量,其中所有尺寸为 <cite>1</cite> 的input尺寸均被删除。

例如,如果<cite>输入</cite>的形状为: 1f976021505083151a3b5d3311ab04c2  ,则张量中的<cite>张量将为: 7af8285a40441ae3080550e9267b63f8。</cite>

给定dim时,仅在给定尺寸上执行挤压操作。 如果<cite>输入</cite>的形状为: squeeze(input, 0)保持张量不变,但是squeeze(input, 1)会将张量压缩为  e48d9b756847043d64d7565153ad4faf形状。

Note

返回的张量与输入张量共享存储,因此更改一个张量的内容将更改另一个张量的内容。

Parameters

  • input (Tensor) – the input tensor.
  • 暗淡的 (python:int , 可选)–如果给定,则仅在此维度上压缩输入
  • out (Tensor, optional) – the output tensor.

Example:

>>> x = torch.zeros(2, 1, 2, 1, 2)
>>> x.size()
torch.Size([2, 1, 2, 1, 2])
>>> y = torch.squeeze(x)
>>> y.size()
torch.Size([2, 2, 2])
>>> y = torch.squeeze(x, 0)
>>> y.size()
torch.Size([2, 1, 2, 1, 2])
>>> y = torch.squeeze(x, 1)
>>> y.size()
torch.Size([2, 2, 1, 2])
torch.stack(tensors, dim=0, out=None) → Tensor¶

将张量的序列沿新维度连接起来。

所有张量都必须具有相同的大小。

Parameters

  • 张量(张量序列)–连接的张量序列
  • 暗淡的 (python:int )–插入的尺寸。 必须介于 0 和级联张量的维数之间(含)
  • out (Tensor, optional) – the output tensor.
torch.t(input) → Tensor¶

期望input为< = 2-D 张量,并转置尺寸 0 和 1。

将按原样返回 0-D 和 1-D 张量,并且可以将 2-D 张量视为transpose(input, 0, 1)的简写函数。

Parameters

input (Tensor) – the input tensor.

Example:

>>> x = torch.randn(())
>>> x
tensor(0.1995)
>>> torch.t(x)
tensor(0.1995)
>>> x = torch.randn(3)
>>> x
tensor([ 2.4320, -0.4608,  0.7702])
>>> torch.t(x)
tensor([.2.4320,.-0.4608,..0.7702])
>>> x = torch.randn(2, 3)
>>> x
tensor([[ 0.4875,  0.9158, -0.5872],
        [ 0.3938, -0.6929,  0.6932]])
>>> torch.t(x)
tensor([[ 0.4875,  0.3938],
        [ 0.9158, -0.6929],
        [-0.5872,  0.6932]])
torch.take(input, index) → Tensor¶

返回给定索引处带有input元素的新张量。 将输入张量视为视为一维张量。 结果采用与索引相同的形状。

Parameters

  • input (Tensor) – the input tensor.
  • 索引 (LongTensor )–张量索引

Example:

>>> src = torch.tensor([[4, 3, 5],
                        [6, 7, 8]])
>>> torch.take(src, torch.tensor([0, 2, 5]))
tensor([ 4,  5,  8])
torch.transpose(input, dim0, dim1) → Tensor¶

返回一个张量,该张量是input的转置版本。 给定的尺寸dim0dim1被交换。

产生的out张量与input张量共享其基础存储,因此更改一个内容将更改另一个内容。

Parameters

  • input (Tensor) – the input tensor.
  • dim0 (python:int )–要转置的第一个维度
  • dim1 (python:int )–要转置的第二维

Example:

>>> x = torch.randn(2, 3)
>>> x
tensor([[ 1.0028, -0.9893,  0.5809],
        [-0.1669,  0.7299,  0.4942]])
>>> torch.transpose(x, 0, 1)
tensor([[ 1.0028, -0.1669],
        [-0.9893,  0.7299],
        [ 0.5809,  0.4942]])
torch.unbind(input, dim=0) → seq¶

删除张量尺寸。

返回给定维度上所有切片的元组,已经没有它。

Parameters

  • 输入 (tensor)–要解除绑定的张量
  • 暗淡的 (python:int )–要移除的尺寸

Example:

>>> torch.unbind(torch.tensor([[1, 2, 3],
>>>                            [4, 5, 6],
>>>                            [7, 8, 9]]))
(tensor([1, 2, 3]), tensor([4, 5, 6]), tensor([7, 8, 9]))
torch.unsqueeze(input, dim, out=None) → Tensor¶

返回在指定位置插入的尺寸为 1 的新张量。

返回的张量与此张量共享相同的基础数据。

可以使用[-input.dim() - 1, input.dim() + 1)范围内的dim值。 负的dim对应于dim = dim + input.dim() + 1处应用的  unsqueeze()  。

Parameters

  • input (Tensor) – the input tensor.
  • 暗淡的 (python:int )–插入单例尺寸的索引
  • out (Tensor, optional) – the output tensor.

Example:

>>> x = torch.tensor([1, 2, 3, 4])
>>> torch.unsqueeze(x, 0)
tensor([[ 1,  2,  3,  4]])
>>> torch.unsqueeze(x, 1)
tensor([[ 1],
        [ 2],
        [ 3],
        [ 4]])
torch.where()¶
torch.where(condition, x, y) → Tensor

返回从xy中选择的元素的张量,具体取决于condition

该操作定义为:

12ee903e238f296681b1cef26fef0f8f

Note

张量conditionxy必须是可广播的。

Parameters

  • 条件 (BoolTensor)–当为 True(非零)时,产生 x,否则产生 y
  • x (tensor)–在conditionTrue的索引处选择的值
  • y (tensor)–在conditionFalse的索引处选择的值

Returns

形状张量等于conditionxy的广播形状

Return type

Tensor

Example:

>>> x = torch.randn(3, 2)
>>> y = torch.ones(3, 2)
>>> x
tensor([[-0.4620,  0.3139],
        [ 0.3898, -0.7197],
        [ 0.0478, -0.1657]])
>>> torch.where(x > 0, x, y)
tensor([[ 1.0000,  0.3139],
        [ 0.3898,  1.0000],
        [ 0.0478,  1.0000]])
torch.where(condition) → tuple of LongTensor

torch.where(condition)torch.nonzero(condition, as_tuple=True)相同。

Note

另请参见 torch.nonzero() 。

发电机

class torch._C.Generator(device='cpu') → Generator¶

创建并返回一个生成器对象,该对象管理产生伪随机数的算法的状态。 在许多就地随机采样函数中用作关键字参数。

Parameters

设备(torch.device,可选)–生成器所需的设备。

Returns

一个 torch.Generator 对象。

Return type

生成器

Example:

>>> g_cpu = torch.Generator()
>>> g_cuda = torch.Generator(device='cuda')
device¶

Generator.device->设备

获取生成器的当前设备。

Example:

>>> g_cpu = torch.Generator()
>>> g_cpu.device
device(type='cpu')
get_state() → Tensor¶

返回生成器状态为torch.ByteTensor

Returns

一个torch.ByteTensor,其中包含将生成器还原到特定时间点的所有必要位。

Return type

Tensor

Example:

>>> g_cpu = torch.Generator()
>>> g_cpu.get_state()
initial_seed() → int¶

返回用于生成随机数的初始种子。

Example:

>>> g_cpu = torch.Generator()
>>> g_cpu.initial_seed()
2147483647
manual_seed(seed) → Generator¶

设置用于生成随机数的种子。 返回一个<cite>torch.生成器</cite>对象。 建议设置一个大种子,即一个具有 0 和 1 位平衡的数字。 避免在种子中包含许多 0 位。

Parameters

种子 (python:int )–所需的种子。

Returns

An torch.Generator object.

Return type

Generator

Example:

>>> g_cpu = torch.Generator()
>>> g_cpu.manual_seed(2147483647)
seed() → int¶

从 std :: random_device 或当前时间获取不确定的随机数,并将其用作生成器的种子。

Example:

>>> g_cpu = torch.Generator()
>>> g_cpu.seed()
1516516984916
set_state(new_state) → void¶

设置生成器状态。

Parameters

new_state (Torch.ByteTensor )–所需状态。

Example:

>>> g_cpu = torch.Generator()
>>> g_cpu_other = torch.Generator()
>>> g_cpu.set_state(g_cpu_other.get_state())

随机抽样

torch.seed()¶

将用于生成随机数的种子设置为不确定的随机数。 返回用于播种 RNG 的 64 位数字。

torch.manual_seed(seed)¶

设置用于生成随机数的种子。 返回一个<cite>torch.生成器</cite>对象。

Parameters

seed (python:int) – The desired seed.

torch.initial_seed()¶

返回长为 Python <cite>long</cite> 的用于生成随机数的初始种子。

torch.get_rng_state()¶

以 <cite>torch.ByteTensor</cite> 的形式返回随机数生成器状态。

torch.set_rng_state(new_state)¶

设置随机数生成器状态。

Parameters

new_state (torch.ByteTensor )–所需状态

torch.default_generator Returns the default CPU torch.Generator¶
torch.bernoulli(input, *, generator=None, out=None) → Tensor¶

从伯努利分布中提取二进制随机数(0 或 1)。

input张量应为包含用于绘制二进制随机数的概率的张量。 因此,input中的所有值都必须在以下范围内: b96a669c7fd5b0f5ea99f4373d6eab2e  。

输出张量的  55a50814a1cfd1ab95d71ca3a7f311fb  元素将根据input中给出的  55a50814a1cfd1ab95d71ca3a7f311fb概率值绘制一个  21c5bc8d40ee5ab2e18d64aaba8359c7值。

f1fc132b2ecf782642591904c4f9b7de

返回的out张量仅具有值 0 或 1,并且具有与input相同的形状。

out可以具有整数dtype,但是input必须具有浮点dtype

Parameters

  • 输入 (tensor)–伯努利分布的概率值的输入张量
  • 生成器(torch.Generator,可选)–用于采样的伪随机数生成器
  • out (Tensor, optional) – the output tensor.

Example:

>>> a = torch.empty(3, 3).uniform_(0, 1)  # generate a uniform random matrix with range [0, 1]
>>> a
tensor([[ 0.1737,  0.0950,  0.3609],
        [ 0.7148,  0.0289,  0.2676],
        [ 0.9456,  0.8937,  0.7202]])
>>> torch.bernoulli(a)
tensor([[ 1.,  0.,  0.],
        [ 0.,  0.,  0.],
        [ 1.,  1.,  1.]])

>>> a = torch.ones(3, 3) # probability of drawing "1" is 1
>>> torch.bernoulli(a)
tensor([[ 1.,  1.,  1.],
        [ 1.,  1.,  1.],
        [ 1.,  1.,  1.]])
>>> a = torch.zeros(3, 3) # probability of drawing "1" is 0
>>> torch.bernoulli(a)
tensor([[ 0.,  0.,  0.],
        [ 0.,  0.,  0.],
        [ 0.,  0.,  0.]])
torch.multinomial(input, num_samples, replacement=False, *, generator=None, out=None) → LongTensor¶

返回一个张量,其中每行包含num_samples索引,这些索引是从位于张量input的相应行中的多项式概率分布中采样的。

Note

input的行不需要加总为 1(在这种情况下,我们将这些值用作权重),但必须为非负数,有限且总和为非零。

根据每个样本的采样时间,索引从左到右排序(第一个样本放在第一列中)。

如果input是向量,则out是大小num_samples的向量。

如果input是具有 <cite>m</cite> 行的矩阵,则out是形状 5d2de3653458e6a14a8a4f0fa87b3ad1的矩阵。

如果替换为True,则抽取样本进行替换。

如果没有,则它们将被替换而不会被绘制,这意味着当为一行绘制样本索引时,无法为该行再次绘制它。

Note

如果绘制时不进行替换,则num_samples必须小于input中非零元素的数目(如果是矩阵,则必须小于input每行中非零元素的最小数目)。

Parameters

  • 输入 (tensor)–包含概率的输入张量
  • num_samples (python:int )–要绘制的样本数
  • 替换 (bool , 可选)–是否使用替换绘制
  • generator  (torch.Generator, optional) – a pseudorandom number generator for sampling
  • out (Tensor, optional) – the output tensor.

Example:

>>> weights = torch.tensor([0, 10, 3, 0], dtype=torch.float) # create a tensor of weights
>>> torch.multinomial(weights, 2)
tensor([1, 2])
>>> torch.multinomial(weights, 4) # ERROR!
RuntimeError: invalid argument 2: invalid multinomial distribution (with replacement=False,
not enough non-negative category to sample) at ../aten/src/TH/generic/THTensorRandom.cpp:320
>>> torch.multinomial(weights, 4, replacement=True)
tensor([ 2,  1,  1,  1])
torch.normal()¶
torch.normal(mean, std, *, generator=None, out=None) → Tensor

返回从均值和标准差给出的独立正态分布中得出的随机数张量。

mean 是一个张量,每个输出元素的正态分布均值

std 是一个张量,每个输出元素的正态分布的标准偏差

mean 和 std 的形状不需要匹配,但是每个张量中元素的总数必须相同。

Note

当形状不匹配时,将 mean 的形状用作返回的输出张量的形状

Parameters

  • 均值 (tensor)–每个元素均值的张量
  • std (tensor)–每个元素的标准偏差张量
  • generator  (torch.Generator, optional) – a pseudorandom number generator for sampling
  • out (Tensor, optional) – the output tensor.

Example:

>>> torch.normal(mean=torch.arange(1., 11.), std=torch.arange(1, 0, -0.1))
tensor([  1.0425,   3.5672,   2.7969,   4.2925,   4.7229,   6.2134,
          8.0505,   8.1408,   9.0563,  10.0566])
torch.normal(mean=0.0, std, out=None) → Tensor

与上面的功能相似,但均值在所有绘制的元素之间共享。

Parameters

  • 平均值 (python:float , 可选)–所有分布的平均值
  • std (Tensor) – the tensor of per-element standard deviations
  • out (Tensor, optional) – the output tensor.

Example:

>>> torch.normal(mean=0.5, std=torch.arange(1., 6.))
tensor([-1.2793, -1.0732, -2.0687,  5.1177, -1.2303])
torch.normal(mean, std=1.0, out=None) → Tensor

与上面的函数相似,但是标准偏差在所有绘制的元素之间共享。

Parameters

  • mean (Tensor) – the tensor of per-element means
  • std (python:float , 可选)–所有发行版的标准差
  • out (tensor , 可选)–输出张量

Example:

>>> torch.normal(mean=torch.arange(1., 6.))
tensor([ 1.1552,  2.6148,  2.6535,  5.8318,  4.2361])
torch.normal(mean, std, size, *, out=None) → Tensor

与上述功能相似,但均值和标准差在所有绘制的元素之间共享。  所得张量的大小由size给出。

Parameters

  • 平均值 (python:float )–所有分布的平均值
  • std (python:float )–所有分布的标准偏差
  • 大小 (python:int ... )–定义输出张量形状的整数序列。
  • out (Tensor, optional) – the output tensor.

Example:

>>> torch.normal(2, 3, size=(1, 4))
tensor([[-1.3987, -1.9544,  3.6048,  0.7909]])
torch.rand(*size, out=None, dtype=None, layout=torch.strided, device=None, requires_grad=False) → Tensor¶

从区间  22e975ba7fcff9173338360452b96179返回均匀分布的随机张量

张量的形状由变量参数size定义。

Parameters

  • size (python:int...) – a sequence of integers defining the shape of the output tensor. Can be a variable number of arguments or a collection like a list or tuple.
  • out (Tensor, optional) – the output tensor.
  • dtype (torch.dtype, optional) – the desired data type of returned tensor. Default: if None, uses a global default (see  torch.set_default_tensor_type() ).
  • layout (torch.layout, optional) – the desired layout of returned Tensor. Default: torch.strided.
  • device  (torch.device, optional) – the desired device of returned tensor. Default: if None, uses the current device for the default tensor type (see torch.set_default_tensor_type()). device will be the CPU for CPU tensor types and the current CUDA device for CUDA tensor types.
  • requires_grad (bool__, optional) – If autograd should record operations on the returned tensor. Default: False.

Example:

>>> torch.rand(4)
tensor([ 0.5204,  0.2503,  0.3525,  0.5673])
>>> torch.rand(2, 3)
tensor([[ 0.8237,  0.5781,  0.6879],
        [ 0.3816,  0.7249,  0.0998]])
torch.rand_like(input, dtype=None, layout=None, device=None, requires_grad=False) → Tensor¶

返回与input大小相同的张量,该张量由间隔 上均匀分布的随机数填充。 torch.rand_like(input)等效于torch.rand(input.size(), dtype=input.dtype, layout=input.layout, device=input.device)

Parameters

  • input (Tensor) – the size of input will determine size of the output tensor.
  • dtype (torch.dtype, optional) – the desired data type of returned Tensor. Default: if None, defaults to the dtype of input.
  • layout (torch.layout, optional) – the desired layout of returned tensor. Default: if None, defaults to the layout of input.
  • device(torch.device, optional) – the desired device of returned tensor. Default: if None, defaults to the device of input.
  • requires_grad (bool__, optional) – If autograd should record operations on the returned tensor. Default: False.
torch.randint(low=0, high, size, *, generator=None, out=None, dtype=None, layout=torch.strided, device=None, requires_grad=False) → Tensor¶

返回一个由在low(包括)和high(不包括)之间均匀生成的随机整数填充的张量。

The shape of the tensor is defined by the variable argument size.

Parameters

  • 低 (python:int , 可选)–从分布中得出的最低整数。 默认值:0
  • 高 (python:int )–从分布中得出的最高整数之上一个。
  • 大小(元组)–定义输出张量形状的元组。
  • generator (torch.Generator, optional) – a pseudorandom number generator for sampling
  • out (Tensor, optional) – the output tensor.
  • dtype (torch.dtype, optional) – the desired data type of returned tensor. Default: if None, uses a global default (see torch.set_default_tensor_type()).
  • layout (torch.dtype, optional) – the desired data type of returned tensor. Default: if None, uses a global default (see  torch.set_default_tensor_type() ).
  • device  (torch.device, optional) – the desired device of returned tensor. Default: if None, uses the current device for the default tensor type (see torch.set_default_tensor_type()). device will be the CPU for CPU tensor types and the current CUDA device for CUDA tensor types.
  • requires_grad (bool__, optional) – If autograd should record operations on the returned tensor. Default: False.

Example:

>>> torch.randint(3, 5, (3,))
tensor([4, 3, 4])

>>> torch.randint(10, (2, 2))
tensor([[0, 2],
        [5, 5]])

>>> torch.randint(3, 10, (2, 2))
tensor([[4, 5],
        [6, 7]])
torch.randint_like(input, low=0, high, dtype=None, layout=torch.strided, device=None, requires_grad=False) → Tensor¶

返回具有与张量input相同形状的张量,其中填充了在low(包括)和high(排除)之间均匀生成的随机整数。

Parameters

  • input (Tensor) – the size of input will determine size of the output tensor.
  • low (python:int__, optional) – Lowest integer to be drawn from the distribution. Default: 0.
  • high (python:int) – One above the highest integer to be drawn from the distribution.
  • dtype  (torch.dtype, optional) – the desired data type of returned Tensor. Default: if None, defaults to the dtype of input.
  • layout (torch.layout, optional) – the desired layout of returned tensor. Default: if None, defaults to the layout of input.
  • device (torch.device, optional) – the desired device of returned tensor. Default: if None, defaults to the device of input.
  • requires_grad (bool__, optional) – If autograd should record operations on the returned tensor. Default: False.
torch.randn(*size, out=None, dtype=None, layout=torch.strided, device=None, requires_grad=False) → Tensor¶

从平均值为 <cite>0</cite> ,方差为 <cite>1</cite> 的正态分布中返回一个填充有随机数的张量(也称为标准正态分布)。

6a5aeab1deaf496af03eb65c0690d32b

The shape of the tensor is defined by the variableargument size.

Parameters

  • size (python:int...) – a sequence of integers defining the shape of the output tensor. Can be a variable number of arguments or a collection like a list or tuple.
  • out (Tensor, optional) – the output tensor.
  • dtype (torch.dtype, optional) – the desired data type of returned tensor. Default: if None, uses a global default (see  torch.set_default_tensor_type() ).
  • layout (torch.layout, optional) – the desired layout of returned Tensor. Default: torch.strided.
  • device (torch.device, optional) – the desired device of returned tensor. Default: if None, uses the current device for the default tensor type (see torch.set_default_tensor_type()). device will be the CPU for CPU tensor types and the current CUDA device for CUDA tensor types.
  • requires_grad (bool__, optional) – If autograd should record operations on the returned tensor.Default: False.

Example:

>>> torch.randn(4)
tensor([-2.1436,  0.9966,  2.3426, -0.6366])
>>> torch.randn(2, 3)
tensor([[ 1.5954,  2.8929, -1.0923],
        [ 1.1719, -0.4709, -0.1996]])
torch.randn_like(input, dtype=None, layout=None, device=None, requires_grad=False) → Tensor¶

返回一个与input相同大小的张量,该张量由均值 0 和方差 1 的正态分布的随机数填充。torch.randn_like(input)等效于torch.randn(input.size(), dtype=input.dtype, layout=input.layout, device=input.device)

Parameters

  • input (Tensor) – the size of input will determine size of the output tensor.
  • dtype (torch.dtype,  optional) – the desired data type of returned Tensor. Default: if None, defaults to the dtype of input.
  • layout (torch.layout, optional) – the desired layout of returned tensor. Default: if None, defaults to the layout of input.
  • device (torch.device, optional) – the desired device of returned tensor. Default: if None, defaults to the device of input.
  • requires_grad (bool__, optional) – If autograd should record operations on the returned tensor. Default: False.
torch.randperm(n, out=None, dtype=torch.int64, layout=torch.strided, device=None, requires_grad=False) → LongTensor¶

返回从0n - 1的整数的随机排列。

Parameters

  • n (python:int )–上限(不包括)
  • out (Tenso, optional) – the output tensor.
  • dtype(torch.dtype ,可选)–返回张量的所需数据类型。 默认值:torch.int64
  • layout (torch.layout, optional) – the desired layout of returned Tensor. Default: torch.strided.
  • device (torch.device, optional) – the desired device of returned tensor. Default: if None, uses the current device for the default tensor type (see torch.set_default_tensor_type()). device will be the CPU for CPU tensor types and the current CUDA device for CUDA tensor types.
  • requires_grad (bool__, optional) – If autograd should record operations on the returned tensor. Default: False.

Example:

>>> torch.randperm(4)
tensor([2, 1, 0, 3])

就地随机抽样

在 Tensor 上还定义了一些就地随机采样函数。 单击以查看其文档:

  • torch.Tensor.bernoulli_() - torch.bernoulli() 的就地版本
  • torch.Tensor.cauchy_() -从柯西分布中得出的数字
  • torch.Tensor.exponential_() -从指数分布中得出的数字
  • torch.Tensor.geometric_() -从几何分布中绘制的元素
  • torch.Tensor.log_normal_()  -来自对数正态分布的样本
  • torch.Tensor.normal_() - torch.normal() 的就地版本
  • torch.Tensor.random_() -从离散均匀分布中采样的数字
  • torch.Tensor.uniform_() -从连续均匀分布中采样的数字

准随机抽样

class torch.quasirandom.SobolEngine(dimension, scramble=False, seed=None)¶

torch.quasirandom.SobolEngine 是用于生成(加扰)Sobol 序列的引擎。 Sobol 序列是低差异准随机序列的一个示例。

用于 Sobol 序列的引擎的这种实现方式能够对最大维度为 1111 的序列进行采样。它使用方向编号生成这些序列,并且这些编号已从此处改编而来。

参考文献

  • Art B. Owen。 争夺 Sobol 和 Niederreiter-Xing 点。 复杂性杂志,14(4):466-489,1998 年 12 月。
  • I. M. Sobol。 立方体中点的分布和积分的准确评估。 嗯 Vychisl。 垫。 我在。 Phys。,7:784-802,1967。

Parameters

  • 尺寸 (Int )–要绘制的序列的尺寸
  • 扰乱 (bool , 可选)–将其设置为True将产生扰乱的 Sobol 序列。 加扰能够产生更好的 Sobol 序列。默认值:False
  • 种子 (Int , 可选)–这是加扰的种子。 如果指定,则将随机数生成器的种子设置为此。 否则,它将使用随机种子。默认值:None

例子:

>>> soboleng = torch.quasirandom.SobolEngine(dimension=5)
>>> soboleng.draw(3)
tensor([[0.5000, 0.5000, 0.5000, 0.5000, 0.5000],
        [0.7500, 0.2500, 0.7500, 0.2500, 0.7500],
        [0.2500, 0.7500, 0.2500, 0.7500, 0.2500]])
draw(n=1, out=None, dtype=torch.float32)¶

从 Sobol 序列中绘制n点序列的功能。 请注意,样本取决于先前的样本。 结果的大小为 4a928fac0225cfecb1270d6afba7caf9

Parameters

  • n (Int , 可选)–绘制点序列的长度。 默认值:1
  • out (tensor , 可选)–输出张量
  • dtype (torch.dtype,可选)–返回的张量的所需数据类型。 默认值:torch.float32
fast_forward(n)¶

通过n步骤快速前进SobolEngine状态的功能。 这等效于不使用样本绘制n样本。

Parameters

n (Int )–快进的步数。

reset()¶

SobolEngine重置为基本状态的功能。

序列化

torch.save(obj, f, pickle_module=<module 'pickle' from '/opt/conda/lib/python3.6/pickle.py'>, pickle_protocol=2, _use_new_zipfile_serialization=False)¶

将对象保存到磁盘文件。

另请参见:推荐的模型保存方法

Parameters

  • obj –保存的对象
  • f –类似于文件的对象(必须实现写入和刷新)或包含文件名的字符串
  • pickle_module –用于腌制元数据和对象的模块
  • pickle_protocol –可以指定为覆盖默认协议

Warning

如果使用的是 Python 2,则 torch.save() 不支持StringIO.StringIO作为有效的类似文件的对象。 这是因为 write 方法应返回写入的字节数; StringIO.write()不这样做。

请改用io.BytesIO之类的东西。

>>> # Save to file
>>> x = torch.tensor([0, 1, 2, 3, 4])
>>> torch.save(x, 'tensor.pt')
>>> # Save to io.BytesIO buffer
>>> buffer = io.BytesIO()
>>> torch.save(x, buffer)
torch.load(f, map_location=None, pickle_module=<module 'pickle' from '/opt/conda/lib/python3.6/pickle.py'>, **pickle_load_args)¶

从文件加载用 torch.save() 保存的对象。

torch.load() 使用 Python 的解开工具,但会特别处理位于张量之下的存储。 它们首先在 CPU 上反序列化,然后移到保存它们的设备上。 如果失败(例如,因为运行时系统没有某些设备),则会引发异常。 但是,可以使用 map_location参数将存储动态重新映射到一组备用设备。

如果map_location是可调用的,则将为每个序列化存储调用一次,并带有两个参数:storage 和 location。 storage 参数将是驻留在 CPU 上的存储的初始反序列化。 每个序列化存储都有一个与之关联的位置标签,该标签标识了从中进行保存的设备,该标签是传递给map_location的第二个参数。 内置位置标签是用于 CPU 张量的'cpu'和用于 CUDA 张量的'cuda:device_id'(例如'cuda:2')。 map_location应该返回None或存储。 如果map_location返回存储,它将用作最终反序列化的对象,已经移至正确的设备。 否则, torch.load() 将退回到默认行为,就像未指定map_location一样。

如果map_location是 torch.device 对象或与设备标签冲突的字符串,则它指示应加载所有张量的位置。

否则,如果map_location是字典,它将用于将文件(键)中出现的位置标签重新映射到指定将存储位置(值)放置的位置标签。

用户扩展可以使用torch.serialization.register_package()注册自己的位置标签以及标记和反序列化方法。

Parameters

  • f –类似于文件的对象(必须实现read(),:methreadline,:methtell和:methseek)或包含文件名的字符串
  • map_location –函数, torch.device ,字符串或指定如何重新映射存储位置的字典
  • pickle_module –用于解开元数据和对象的模块(必须与用于序列化文件的pickle_module匹配)
  • pickle_load_args –(仅适用于 Python 3)可选关键字参数传递给pickle_module.load()pickle_module.Unpickler(),例如errors=...

Note

当您在包含 GPU 张量的文件上调用 torch.load() 时,这些张量将默认加载到 GPU。 您可以先调用torch.load(.., map_location='cpu'),然后再调用load_state_dict(),以避免在加载模型检查点时 GPU RAM 激增。

Note

默认情况下,我们将字节字符串解码为utf-8。 这是为了避免在 Python 3 中加载 Python 2 保存的文件时出现常见错误情况UnicodeDecodeError: 'ascii' codec can't decode byte 0x...。如果此默认设置不正确,则可以使用额外的encoding关键字参数来指定应如何加载这些对象,例如encoding='latin1'使用latin1编码将它们解码为字符串,encoding='bytes'将它们保留为字节数组,以后可以使用byte_array.decode(...)进行解码。

Example

>>> torch.load('tensors.pt')
## Load all tensors onto the CPU
>>> torch.load('tensors.pt', map_location=torch.device('cpu'))
## Load all tensors onto the CPU, using a function
>>> torch.load('tensors.pt', map_location=lambda storage, loc: storage)
## Load all tensors onto GPU 1
>>> torch.load('tensors.pt', map_location=lambda storage, loc: storage.cuda(1))
## Map tensors from GPU 1 to GPU 0
>>> torch.load('tensors.pt', map_location={'cuda:1':'cuda:0'})
## Load tensor from io.BytesIO object
>>> with open('tensor.pt', 'rb') as f:
        buffer = io.BytesIO(f.read())
>>> torch.load(buffer)
## Load a module with 'ascii' encoding for unpickling
>>> torch.load('module.pt', encoding='ascii')

平行性

torch.get_num_threads() → int¶

返回用于并行化 CPU 操作的线程数

torch.set_num_threads(int)¶

设置用于 CPU 上的内部运算并行的线程数。 警告:为确保使用正确的线程数,必须在运行 eager,JIT 或 autograd 代码之前调用 set_num_threads。

torch.get_num_interop_threads() → int¶

返回用于 CPU 上的互操作并行的线程数(例如,在 JIT 解释器中)

torch.set_num_interop_threads(int)¶

设置用于 CPU 上的互操作并行性(例如,在 JIT 解释器中)的线程数。 警告:只能在一次操作间并行工作开始之前(例如 JIT 执行)调用一次。

局部禁用梯度计算

上下文管理器torch.no_grad()torch.enable_grad()torch.set_grad_enabled()有助于局部禁用和启用梯度计算。 有关其用法的更多详细信息,请参见局部禁用梯度计算。 这些上下文管理器是线程本地的,因此如果您使用threading模块等将工作发送到另一个线程,它们将无法工作。

Examples:

>>> x = torch.zeros(1, requires_grad=True)
>>> with torch.no_grad():
...     y = x * 2
>>> y.requires_grad
False

>>> is_train = False
>>> with torch.set_grad_enabled(is_train):
...     y = x * 2
>>> y.requires_grad
False

>>> torch.set_grad_enabled(True)  # this can also be used as a function
>>> y = x * 2
>>> y.requires_grad
True

>>> torch.set_grad_enabled(False)
>>> y = x * 2
>>> y.requires_grad
False

数学运算

逐点操作

torch.abs(input, out=None) → Tensor¶

计算给定input张量的按元素的绝对值。

726d369abb9c76e751e626cbfef10220

Parameters

  • input (Tensor) – the input tensor.
  • out (Tensor, optional) – the output tensor.

Example:

>>> torch.abs(torch.tensor([-1, -2, 3]))
tensor([ 1,  2,  3])
torch.acos(input, out=None) → Tensor¶

返回带有input元素的反余弦的新张量。

25323a5363649a36549d717d5a3cbc3e

Parameters

  • input (Tensor) – the input tensor.
  • out (Tensor, optional) – the output tensor.

Example:

>>> a = torch.randn(4)
>>> a
tensor([ 0.3348, -0.5889,  0.2005, -0.1584])
>>> torch.acos(a)
tensor([ 1.2294,  2.2004,  1.3690,  1.7298])
torch.add()¶
torch.add(input, other, out=None)

将标量other添加到输入input的每个元素中,并返回一个新的结果张量。

1e18f010799098c127bab9465476d1fb

如果input的类型为 FloatTensor 或 DoubleTensor,则other必须为实数,否则应为整数。

Parameters

  • input (Tensor) – the input tensor.
  • 值(编号)–要添加到input每个元素的编号
Keyword Arguments

out (Tensor, optional) – the output tensor.

Example:

>>> a = torch.randn(4)
>>> a
tensor([ 0.0202,  1.0985,  1.3506, -0.6056])
>>> torch.add(a, 20)
tensor([ 20.0202,  21.0985,  21.3506,  19.3944])
torch.add(input, alpha=1, other, out=None)

张量other的每个元素乘以标量alpha,然后加到张量input的每个元素上。 返回结果张量。

inputother的状必须是可广播的。

5a5bc06d446342edf134fb0e87f755ec

如果other的类型为 FloatTensor 或 DoubleTensor,则alpha必须为实数,否则应为整数。

Parameters

  • 输入 (tensor)–第一个输入张量
  • alpha (数字)– other的标量乘法器
  • 其他 (tensor)–第二个输入张量
Keyword Arguments

out (Tensor, optional) – the output tensor.

Example:

>>> a = torch.randn(4)
>>> a
tensor([-0.9732, -0.3497,  0.6245,  0.4022])
>>> b = torch.randn(4, 1)
>>> b
tensor([[ 0.3743],
        [-1.7724],
        [-0.5811],
        [-0.8017]])
>>> torch.add(a, 10, b)
tensor([[  2.7695,   3.3930,   4.3672,   4.1450],
        [-18.6971, -18.0736, -17.0994, -17.3216],
        [ -6.7845,  -6.1610,  -5.1868,  -5.4090],
        [ -8.9902,  -8.3667,  -7.3925,  -7.6147]])
torch.addcdiv(input, value=1, tensor1, tensor2, out=None) → Tensor¶

执行tensor1除以tensor2的元素,将结果乘以标量value并将其加到input上。

69053b1e1ce2faac8aae0ee4370775b9

inputtensor1tensor2的形状必须是可广播。

对于类型为 <cite>FloatTensor</cite> 或 <cite>DoubleTensor</cite> 的输入,value必须为实数,否则为整数。

Parameters

  • 输入 (tensor)–要添加的张量
  • 值(编号 , 可选)–  的乘数
  • 张量 1 (tensor)–分子张量
  • 张量 2 (tensor)–分母张量
  • out (Tensor, optional) – the output tensor.

Example:

>>> t = torch.randn(1, 3)
>>> t1 = torch.randn(3, 1)
>>> t2 = torch.randn(1, 3)
>>> torch.addcdiv(t, 0.1, t1, t2)
tensor([[-0.2312, -3.6496,  0.1312],
        [-1.0428,  3.4292, -0.1030],
        [-0.5369, -0.9829,  0.0430]])
torch.addcmul(input, value=1, tensor1, tensor2, out=None) → Tensor¶

tensor1tensor2进行元素逐项乘法,将结果与标量value相乘,然后将其与input相加。

6bfdc94f1a8a13ee1a6ebc9af2d74fda

tensor ,tensor1tensor2的形状必须是可广播的。

For inputs of type <cite>FloatTensor</cite> or <cite>DoubleTensor</cite>, value must  be a real number, otherwise an integer.

Parameters

  • input (Tensor) – the tensor to be added
  • 值(编号 , 可选)–  的乘数
  • 张量 1 (tensor)–要相乘的张量
  • 张量 2 (tensor)–要相乘的张量
  • out (Tensor, optional) – the output tensor.

Example:

>>> t = torch.randn(1, 3)
>>> t1 = torch.randn(3, 1)
>>> t2 = torch.randn(1, 3)
>>> torch.addcmul(t, 0.1, t1, t2)
tensor([[-0.8635, -0.6391,  1.6174],
        [-0.7617, -0.5879,  1.7388],
        [-0.8353, -0.6249,  1.6511]])
torch.angle(input, out=None) → Tensor¶

计算给定input张量的元素方向角(以弧度为单位)。

5af8c34b3d4c490c508caf11a458d5d6

Parameters

  • input (Tensor) – the input tensor.
  • out (Tensor, optional) – the output tensor.

Example:

>>> torch.angle(torch.tensor([-1 + 1j, -2 + 2j, 3 - 3j]))*180/3.14159
tensor([ 135.,  135,  -45])
torch.asin(input, out=None) → Tensor¶

返回带有input元素的反正弦值的新张量。

fa8ed00ae85106eea8e8a11b6b66d898

Parameters

  • input (Tensor) – the input tensor.
  • out (Tensor, optional) – the output tensor.

Example:

>>> a = torch.randn(4)
>>> a
tensor([-0.5962,  1.4985, -0.4396,  1.4525])
>>> torch.asin(a)
tensor([-0.6387,     nan, -0.4552,     nan])
torch.atan(input, out=None) → Tensor¶

返回带有input元素的反正切的新张量。

8614369d5e7413ce21e9a5d6bcf786cb

Parameters

  • input (Tensor) – the input tensor.
  • out (Tensor, optional) – the output tensor.

Example:

>>> a = torch.randn(4)
>>> a
tensor([ 0.2341,  0.2539, -0.6256, -0.6448])
>>> torch.atan(a)
tensor([ 0.2299,  0.2487, -0.5591, -0.5727])
torch.atan2(input, other, out=None) → Tensor¶

考虑象限的  c45e59e7ab5c2f7cff1ff37748c0d62b元素逐级反正切。 返回一个新的张量,其矢量  2cb0a971340ffd46996a0cf9eb81d376与矢量  8456c6ac83b24dbe917ff5a29a771bd7之间的弧度为符号角。 (请注意,第二个参数  4e409083ee35ea2eeab094ea4f9bf730是 x 坐标,而第一个参数  0c7468da87ed4da7f144e93eb6e7e60e是 y 坐标。)

iinputother的形状必须是可广播的。

Parameters

  • input (Tenso) – the first input tensor
  • other (Tensor) – the second input tensor
  • out (Tensor, optional) – the output tensor.

Example:

>>> a = torch.randn(4)
>>> a
tensor([ 0.9041,  0.0196, -0.3108, -2.4423])
>>> torch.atan2(a, torch.randn(4))
tensor([ 0.9833,  0.0811, -1.9743, -1.4151])
torch.bitwise_not(input, out=None) → Tensor¶

计算给定输入张量的按位非。 输入张量必须是整数或布尔类型。 对于布尔张量,它计算逻辑非。

Parameters

  • input (Tensor) – the input tensor.
  • out (Tensor, optional) – the output tensor.

Example

>>> torch.bitwise_not(torch.tensor([-1, -2, 3], dtype=torch.int8))
tensor([ 0,  1, -4], dtype=torch.int8)
torch.bitwise_xor(input, other, out=None) → Tensor¶

计算inputother的按位 XOR。 输入张量必须是整数或布尔类型。 对于布尔张量,它计算逻辑 XOR。

Parameters

  • 输入 –第一个输入张量
  • 其他 –第二个输入张量
  • out (Tensor, optional) – the output tensor.

Example

>>> torch.bitwise_xor(torch.tensor([-1, -2, 3], dtype=torch.int8), torch.tensor([1, 0, 3], dtype=torch.int8))
tensor([-2, -2,  0], dtype=torch.int8)
>>> torch.bitwise_xor(torch.tensor([True, True, False]), torch.tensor([False, True, False]))
tensor([ True, False, False])
torch.ceil(input, out=None) → Tensor¶

返回带有input元素的 ceil 的新张量,该元素大于或等于每个元素的最小整数。

3425145dda63005bbc6ff0b2f0331aa7

Parameters

  • input (Tensor) – the input tensor.
  • out (Tensor, optional) – the output tensor.

Example:

>>> a = torch.randn(4)
>>> a
tensor([-0.6341, -1.4208, -1.0900,  0.5826])
>>> torch.ceil(a)
tensor([-0., -1., -1.,  1.])
torch.clamp(input, min, max, out=None) → Tensor¶

input中的所有元素限制在 <cite>[</cite> min , max <cite>]</cite> 范围内,并返回结果张量:

597a5259be5b1a330eaa6858b12b8994

如果input的类型为 <cite>FloatTensor</cite> 或 <cite>DoubleTensor</cite> ,则参数 min 和 max 必须为实数,否则为实数 应该是整数。

Parameters

  • input (Tensor) – the input tensor.
  • min (编号)–要钳制的范围的下限
  • 最大(编号)–要钳位的范围的上限
  • out (Tensor, optional) – the output tensor.

Example:

>>> a = torch.randn(4)
>>> a
tensor([-1.7120,  0.1734, -0.0478, -0.0922])
>>> torch.clamp(a, min=-0.5, max=0.5)
tensor([-0.5000,  0.1734, -0.0478, -0.0922])
torch.clamp(input, *, min, out=None) → Tensor

input中的所有元素限制为大于或等于 min 。

如果input的类型为 <cite>FloatTensor</cite> 或 <cite>DoubleTensor</cite> ,则value应为实数,否则应为整数。

Parameters

  • input (Tensor) – the input tensor.
  • 值(编号)–输出中每个元素的最小值
  • out (Tensor, optional) – the output tensor.

Example:

>>> a = torch.randn(4)
>>> a
tensor([-0.0299, -2.3184,  2.1593, -0.8883])
>>> torch.clamp(a, min=0.5)
tensor([ 0.5000,  0.5000,  2.1593,  0.5000])
torch.clamp(input, *, max, out=None) → Tensor

input中的所有元素限制为小于或等于 <max

If input is of type <cite>FloatTensor</cite> or <cite>DoubleTensor</cite>, value should be a real number, otherwise it should be an integer.

Parameters

  • input (Tensor) – the input tensor.
  • 值(编号)–输出中每个元素的最大值
  • out (Tensor, optional) – the output tensor.

Example:

>>> a = torch.randn(4)
>>> a
tensor([ 0.7753, -0.4702, -0.4599,  1.1899])
>>> torch.clamp(a, max=0.5)
tensor([ 0.5000, -0.4702, -0.4599,  0.5000])
torch.conj(input, out=None) → Tensor¶

计算给定input张量的逐元素共轭。

334473ca71a76cc4b45b56e05f9113be

Parameters

  • input (Tensor) – the input tensor.
  • out (Tensor, optional) – the output tensor.

Example:

>>> torch.conj(torch.tensor([-1 + 1j, -2 + 2j, 3 - 3j]))
tensor([-1 - 1j, -2 - 2j, 3 + 3j])
torch.cos(input, out=None) → Tensor¶

返回带有input元素的余弦的新张量。

a54751040e4e9bbc318ce13170d3575a

Parameters

  • input (Tensor) – the input tensor.
  • out (Tensor, optional) – the output tensor.

Example:

>>> a = torch.randn(4)
>>> a
tensor([ 1.4309,  1.2706, -0.8562,  0.9796])
>>> torch.cos(a)
tensor([ 0.1395,  0.2957,  0.6553,  0.5574])
torch.cosh(input, out=None) → Tensor¶

返回具有input元素的双曲余弦的新张量。

01d5c509dd9312bb70bb293dffb6ee74

Parameters

  • input (Tensor) – the input tensor.
  • out (