Skip to main content
Version: 0.2.1

Interface: Torch

torchlive/torch.Torch

Properties

channelsLast

channelsLast: "channelsLast"


contiguousFormat

contiguousFormat: "contiguousFormat"


double

double: "double"


float

float: "float"


float32

float32: "float32"


float64

float64: "float64"


int

int: "int"


int16

int16: "int16"


int32

int32: "int32"


int64

int64: "int64"


int8

int8: "int8"


jit

jit: JIT

JIT module


long

long: "long"


preserveFormat

preserveFormat: "preserveFormat"


short

short: "short"


uint8

uint8: "uint8"

Methods

arange

arange(end, options?): Tensor

Returns a 1-D tensor of size (end - 0) / 1 with values from the interval [0, end) taken with common difference step beginning from start.

https://pytorch.org/docs/1.12/generated/torch.arange.html

Parameters

NameTypeDescription
endnumberThe ending value for the set of points.
options?TensorOptions

Returns

Tensor

arange(start, end, options?): Tensor

Returns a 1-D tensor of size (end - start) / 1 with values from the interval [start, end) taken with common difference 1 beginning from start.

https://pytorch.org/docs/1.12/generated/torch.arange.html

Parameters

NameTypeDescription
startnumberThe starting value for the set of points.
endnumberThe ending value for the set of points.
options?TensorOptions

Returns

Tensor

arange(start, end, step, options?): Tensor

Returns a 1-D tensor of size (end - start) / step with values from the interval [start, end) taken with common difference step beginning from start.

https://pytorch.org/docs/1.12/generated/torch.arange.html

Parameters

NameTypeDescription
startnumberThe starting value for the set of points.
endnumberThe ending value for the set of points.
stepnumberThe gap between each pair of adjacent points.
options?TensorOptions

Returns

Tensor


cat

cat(tensors, options?): Tensor

Concatenate a list of tensors along the specified axis, which default to be axis 0

https://pytorch.org/docs/1.12/generated/torch.cat.html

Parameters

NameTypeDescription
tensorsTensor[]A sequence of Tensor to be concatenated.
options?Objectused to specify the dimenstion to concate.
options.dim?number-

Returns

Tensor


empty

empty(size, options?): Tensor

Returns a tensor filled with uninitialized data. The shape of the tensor is defined by the variable argument size.

https://pytorch.org/docs/1.12/generated/torch.empty.html

Parameters

NameTypeDescription
sizenumber[]A sequence of integers defining the shape of the output tensor.
options?TensorOptions-

Returns

Tensor


eye

eye(n, m?, options?): Tensor

Returns a tensor filled with ones on the diagonal, and zeroes elsewhere. The shape of the tensor is defined by the arguments n and m.

https://pytorch.org/docs/1.12/generated/torch.eye.html

Parameters

NameTypeDescription
nnumberAn integer defining the number of rows in the result.
m?numberAn integer defining the number of columns in the result. Optional, defaults to n.
options?TensorOptions-

Returns

Tensor


fromBlob

fromBlob(blob, sizes?, options?): Tensor

Exposes the given data as a Tensor without taking ownership of the original data.

note

The function exists in JavaScript and C++ (torch::from_blob).

Parameters

NameTypeDescription
blobanyThe blob holding the data.
sizes?number[]Should specify the shape of the tensor, strides the stride
options?TensorOptionsTensor options in each dimension.

Returns

Tensor


full

full(size, fillValue, options?): Tensor

Creates a tensor of size size filled with fillValue. The tensor’s dtype is default to be torch.float32, unless specified with options.

https://pytorch.org/docs/1.12/generated/torch.full.html

Parameters

NameTypeDescription
sizenumber[]A list of integers defining the shape of the output tensor.
fillValuenumberThe value to fill the output tensor with.
options?TensorOptionsObject to customizing dtype, etc. default to be {dtype: torch.float32}

Returns

Tensor


linspace

linspace(start, end, steps, options?): Tensor

Creates a one-dimensional tensor of size steps whose values are evenly spaced from start to end, inclusive.

https://pytorch.org/docs/1.12/generated/torch.linspace.html

Parameters

NameTypeDescription
startnumberStarting value for the set of points
endnumberEnding value for the set of points
stepsnumberSize of the constructed tensor
options?TensorOptionsObject to customizing dtype. default to be {dtype: torch.float32}

Returns

Tensor


logspace

logspace(start, end, steps, options?): Tensor

Returns a one-dimensional tensor of size steps whose values are evenly spaced from base^start to base^end, inclusive, on a logarithmic scale with base.

https://pytorch.org/docs/1.12/generated/torch.logspace.html

Parameters

NameTypeDescription
startnumberStarting value for the set of points
endnumberEnding value for the set of points
stepsnumberSize of the constructed tensor
options?TensorOptions & { base: number }Object to customizing base and dtype. default to be {base: 10, dtype: torch.float32}

Returns

Tensor


ones

ones(size, options?): Tensor

Returns a tensor filled with the scalar value 1, with the shape defined by the argument size.

https://pytorch.org/docs/1.12/generated/torch.ones.html

Parameters

NameTypeDescription
sizenumber[]A sequence of integers defining the shape of the output tensor.
options?TensorOptionsTensor options.

Returns

Tensor


rand

rand(size, options?): Tensor

Returns a tensor filled with random numbers from a uniform distribution on the interval [0, 1).

Parameters

NameTypeDescription
sizenumber[]A sequence of integers defining the shape of the output tensor.
options?TensorOptionsTensor options.

Returns

Tensor


randint

randint(high, size): Tensor

Returns a tensor filled with random integers generated uniformly between 0 (inclusive) and high (exclusive).

https://pytorch.org/docs/1.12/generated/torch.randint.html

Parameters

NameTypeDescription
highnumberOne above the highest integer to be drawn from the distribution.
sizenumber[]A tuple defining the shape of the output tensor.

Returns

Tensor

randint(low, high, size): Tensor

Returns a tensor filled with random integers generated uniformly between low (inclusive) and high (exclusive).

https://pytorch.org/docs/1.12/generated/torch.randint.html

Parameters

NameTypeDescription
lownumberLowest integer to be drawn from the distribution.
highnumberOne above the highest integer to be drawn from the distribution.
sizenumber[]A tuple defining the shape of the output tensor.

Returns

Tensor


randn

randn(size, options?): Tensor

Returns a tensor filled with random numbers from a normal distribution with mean 0 and variance 1 (also called the standard normal distribution).

https://pytorch.org/docs/1.12/generated/torch.randn.html

Parameters

NameTypeDescription
sizenumber[]A sequence of integers defining the shape of the output tensor.
options?TensorOptionsTensor options.

Returns

Tensor


randperm

randperm(n, options?): Tensor

Returns a random permutation of integers from 0 to n - 1

https://pytorch.org/docs/1.12/generated/torch.randperm.html

Parameters

NameTypeDescription
nnumberThe upper bound (exclusive)
options?TensorOptionsObject to customizing dtype, etc. default to be {dtype: torch.int64}.

Returns

Tensor


tensor

tensor(data, options?): Tensor

Constructs a tensor with no autograd history.

Parameters

NameTypeDescription
datanumber | ItemArrayTensor data as multi-dimensional array.
options?TensorOptionsTensor options.

Returns

Tensor


zeros

zeros(size, options?): Tensor

Returns a tensor filled with the scalar value 0, with the shape defined by the argument size.

https://pytorch.org/docs/1.12/generated/torch.zeros.html

Parameters

NameTypeDescription
sizenumber[]A sequence of integers defining the shape of the output tensor.
options?TensorOptionsTensor options.

Returns

Tensor