torch

package module
v1.11.0-0.1.5 Latest Latest
Warning

This package is not in the latest version of its module.

 Go to latest Published: Apr 2, 2023 License: MIT Imports: 6 Imported by: 6

README

GoTorch

GoTorch Go Report Card

GoTorch is a Golang front-end for pytorch.

Installation

GoTorch requires the installation of libtorch and libcgotorch (a C-binding written for CGo.) To install these libraries to /usr/local, use:

git clone --depth 1 --branch v1.11.0-0.1.4 https://github.com/Kautenja/gotorch /tmp/gotorch
/tmp/gotorch/install.sh
rm -rf /tmp/gotorch

You may refer to the example project for a working demonstration of this library.

Documentation

Index

Constants

This section is empty.

Variables

This section is empty.

Functions

func AllClose 

func AllClose(tensor, other *Tensor, rtol, atol float64) bool

Return true if the two tensors are approximately equal to one another with given relative and absolute tolerances.

This function checks if all `tensor` and `other` satisfy the condition:

|tensor − other| <= atol + rtol * |other|

func Equal 

func Equal(tensor, other *Tensor) bool

Return true if the two tensors are precisely equal element-wise.

func IsComplex 

func IsComplex(tensor *Tensor) bool

Return true if the tensor of a complex data-type.

func IsConj 

func IsConj(tensor *Tensor) bool

Return true if the (complex) tensor is in conjugated form.

func IsDevice 

func IsDevice(deviceName string) bool

Return true if the given device is valid, false otherwise.

func IsFloatingPoint 

func IsFloatingPoint(tensor *Tensor) bool

Return true if the tensor of a floating-point data-type.

func IsGradEnabled 

func IsGradEnabled() bool

Return the global gradient generation state.

func IsNonzero 

func IsNonzero(tensor *Tensor) bool

Return true if the tensor is a single non-zero element (i.e., scalar.)

func ManualSeed 

func ManualSeed(seed int64)

Set the random number generator seed.

func Numel 

func Numel(tensor *Tensor) int64

Return the number of elements in the tensor.

func SetGradEnabled 

func SetGradEnabled(value bool)

Set the global gradient generation state.

func SetNumThreads 

func SetNumThreads(numThreads int32)

Set the number of threads used for intraop parallelism on CPU.

func StdMean 

func StdMean(tensor *Tensor) (*Tensor, *Tensor)

Reduce a tensor to its mean value and standard deviation.

func StdMeanByDim 

func StdMeanByDim(tensor *Tensor, dim int, unbiased, keep_dims bool) (*Tensor, *Tensor)

Reduce a tensor to its mean value and standard deviation along given dimension.

func ToSlice 

func ToSlice(tensor *Tensor) interface{}

Convert a torch Tensor to a Go slice. This function implies a flattening of the tensor to return 1-dimensional vectors.

func TorchVersion 

func TorchVersion() string

Return the version of the libtorch back-end.

func VarMean 

func VarMean(tensor *Tensor) (*Tensor, *Tensor)

Reduce a tensor to its mean value and variance.

func VarMeanByDim 

func VarMeanByDim(tensor *Tensor, dim int, unbiased, keep_dims bool) (*Tensor, *Tensor)

Reduce a tensor to its mean value and variance along given dimension.

func Version 

func Version() string

Return the version of the GoTorch library.

Types

type Device 

type Device struct {
	Pointer C.Device
}

Device wrapper a pointer to C.Device

func NewDevice 

func NewDevice(deviceName string) (device *Device)

Create a new Device.

type Dtype 

type Dtype int8

An enumeration of the data-types available in torch. Typically it would be standard for "Invalid" to correspond to value 0; however, these codes are one-to-one with C++ codes in torch. 

const (
	Byte Dtype = iota
	Char
	Short
	Int
	Long
	Half
	Float
	Double
	ComplexHalf
	ComplexFloat
	ComplexDouble
	Bool
	QInt8
	QUInt8
	QInt32
	BFloat16
	Invalid Dtype = -1
)

func GetDtypeOfKind 

func GetDtypeOfKind(kind reflect.Kind) Dtype

Map an element type kind to its associated Dtype.

func (Dtype) NumBytes 

func (dtype Dtype) NumBytes() int64

Return the number of bytes consumed by each element of the given data-type.

type IValue 

type IValue struct {
	Pointer C.IValue
}

IValue wraps a C.IValue.

func NewIValue 

func NewIValue(data interface{}) (ivalue *IValue)

Create a new IValue from arbitrary data.

func (*IValue) IsBool 

func (ivalue *IValue) IsBool() bool

Return true if the IValue is a boolean.

func (*IValue) IsBoolList 

func (ivalue *IValue) IsBoolList() bool

Return true if the IValue is a list of booleans.

func (*IValue) IsCapsule 

func (ivalue *IValue) IsCapsule() bool

Return true if the IValue is a capsule.

func (*IValue) IsComplexDouble 

func (ivalue *IValue) IsComplexDouble() bool

Return true if the IValue is a complex double-precision floating-point value.

func (*IValue) IsComplexDoubleList 

func (ivalue *IValue) IsComplexDoubleList() bool

Return true if the IValue is a list of complex double-precision floats.

func (*IValue) IsCustomClass 

func (ivalue *IValue) IsCustomClass() bool

Return true if the IValue is a custom class instance.

func (*IValue) IsDevice 

func (ivalue *IValue) IsDevice() bool

Return true if the IValue is a torch device.

func (*IValue) IsDouble 

func (ivalue *IValue) IsDouble() bool

Return true if the IValue is a double-precision floating-point value.

func (*IValue) IsDoubleList 

func (ivalue *IValue) IsDoubleList() bool

Return true if the IValue is a list of double-precision floats.

func (*IValue) IsEnum 

func (ivalue *IValue) IsEnum() bool

Return true if the IValue is an enumeration.

func (*IValue) IsFuture 

func (ivalue *IValue) IsFuture() bool

Return true if the IValue is a future.

func (*IValue) IsGenerator 

func (ivalue *IValue) IsGenerator() bool

Return true if the IValue is a Python generator.

func (*IValue) IsGenericDict 

func (ivalue *IValue) IsGenericDict() bool

Return true if the IValue is a dictionary.

func (*IValue) IsInt 

func (ivalue *IValue) IsInt() bool

Return true if the IValue is an integer.

func (*IValue) IsIntList 

func (ivalue *IValue) IsIntList() bool

Return true if the IValue is a list of integers.

func (*IValue) IsList 

func (ivalue *IValue) IsList() bool

Return true if the IValue is a list.

func (*IValue) IsModule 

func (ivalue *IValue) IsModule() bool

Return true if the IValue is a module.

func (*IValue) IsNil 

func (ivalue *IValue) IsNil() bool

Return true if the IValue references a null pointer.

func (*IValue) IsObject 

func (ivalue *IValue) IsObject() bool

Return true if the IValue is an object.

func (*IValue) IsPtrType 

func (ivalue *IValue) IsPtrType() bool

Return true if the IValue is a void*.

func (*IValue) IsPyObject 

func (ivalue *IValue) IsPyObject() bool

Return true if the IValue is a Python object instance.

func (*IValue) IsQuantizer 

func (ivalue *IValue) IsQuantizer() bool

Return true if the IValue is a quantizer.

func (*IValue) IsRRef 

func (ivalue *IValue) IsRRef() bool

Return true if the IValue is an RRef.

func (*IValue) IsScalar 

func (ivalue *IValue) IsScalar() bool

Return true if the IValue is a scalar.

func (*IValue) IsStorage 

func (ivalue *IValue) IsStorage() bool

Return true if the IValue is a storage medium.

func (*IValue) IsStream 

func (ivalue *IValue) IsStream() bool

Return true if the IValue is a data stream (e.g., a CUDA stream.)

func (*IValue) IsString 

func (ivalue *IValue) IsString() bool

Return true if the IValue is a string.

func (*IValue) IsTensor 

func (ivalue *IValue) IsTensor() bool

Return true if the IValue is a tensor.

func (*IValue) IsTensorList 

func (ivalue *IValue) IsTensorList() bool

Return true if the IValue is a list of tensors.

func (*IValue) IsTuple 

func (ivalue *IValue) IsTuple() bool

Return true if the IValue is a tuple.

func (*IValue) LengthDict 

func (ivalue *IValue) LengthDict() int64

If the ivalue is a dictionary, return the number of items in the dictionary.

func (*IValue) LengthList 

func (ivalue *IValue) LengthList() int64

If the ivalue is a list, return the number of items in the list.

func (*IValue) LengthTuple 

func (ivalue *IValue) LengthTuple() int64

If the ivalue is a tuple, return the number of items in the tuple.

func (*IValue) ToBool 

func (ivalue *IValue) ToBool() bool

Convert the IValue to a boolean.

func (*IValue) ToComplexDouble 

func (ivalue *IValue) ToComplexDouble() complex128

Convert the IValue to a complex double-precision float.

func (*IValue) ToDevice 

func (ivalue *IValue) ToDevice() (device *Device)

Convert the IValue to a torch device.

func (*IValue) ToDouble 

func (ivalue *IValue) ToDouble() float64

Convert the IValue to a double-precision float.

func (*IValue) ToGenericDict 

func (ivalue *IValue) ToGenericDict() map[interface{}]*IValue

Convert the IValue to a generic dictionary of IValues (as a map of interfaces to IValues.)

func (*IValue) ToInt 

func (ivalue *IValue) ToInt() int

Convert the IValue to an integer.

func (*IValue) ToList 

func (ivalue *IValue) ToList() []*IValue

Convert the IValue to a generic list of IValues (as a slice.)

func (*IValue) ToNone 

func (ivalue *IValue) ToNone() string

Convert the IValue to a null pointer (i.e., the string "None".)

func (*IValue) ToString 

func (ivalue *IValue) ToString() string

Convert the IValue to a string.

func (*IValue) ToTensor 

func (ivalue *IValue) ToTensor() *Tensor

Convert the IValue to a tensor.

func (*IValue) ToTensorList 

func (ivalue *IValue) ToTensorList() []*Tensor

Convert the IValue to a list of tensors.

func (*IValue) ToTuple 

func (ivalue *IValue) ToTuple() []*IValue

Convert the IValue to a generic tuple of IValues (as a slice.)

type Tensor 

type Tensor struct {
	Pointer C.Tensor
}

Tensor wraps a pointer to a C.Tensor as an unsafe Pointer.

func Abs 

func Abs(tensor *Tensor) *Tensor

Take the absolute value of input.

func Abs_ 

func Abs_(tensor *Tensor) *Tensor

In-place version of Abs().

func Add 

func Add(tensor, other *Tensor, alpha float32) *Tensor

Adds other, scaled by alpha, to input.

func All 

func All(tensor *Tensor) *Tensor

Check if all values in the tensor evaluate to true.

func AllByDim 

func AllByDim(tensor *Tensor, dim int, keep_dims bool) *Tensor

Check if all values in the tensor evaluate to true along the given dimension.

func Any 

func Any(tensor *Tensor) *Tensor

Check if any values in the tensor evaluate to true.

func AnyByDim 

func AnyByDim(tensor *Tensor, dim int, keep_dims bool) *Tensor

Check if any values in the tensor evaluate to true along the given dimension.

func Arange 

func Arange(begin, end, step float32, options *TensorOptions) *Tensor

Create an inclusive range tensor from begin position to end position along integer step size.

func Argmax 

func Argmax(tensor *Tensor) *Tensor

Reduce a tensor to its maximum index.

func ArgmaxByDim 

func ArgmaxByDim(tensor *Tensor, dim int, keep_dims bool) *Tensor

Reduce a tensor to its maximum index along the given dimension.

func Argmin 

func Argmin(tensor *Tensor) *Tensor

Reduce a tensor to its minimum index.

func ArgminByDim 

func ArgminByDim(tensor *Tensor, dim int, keep_dims bool) *Tensor

Reduce a tensor to its minimum index along the given dimension.

func Cat 

func Cat(tensors []*Tensor, dim int64) *Tensor

Concatenate the given sequence of tensors in the given dimension. All tensors must either have the same shape (except in the concatenating dimension) or be empty.

func Clamp 

func Clamp(tensor, minimum, maximum *Tensor) *Tensor

Clamps all elements in input into the range [min, max].

func ClampMax 

func ClampMax(tensor, maximum *Tensor) *Tensor

Clamps all elements in input to have maximum value of input tensor.

func ClampMax_ 

func ClampMax_(tensor, maximum *Tensor) *Tensor

In-place version of torch.ClampMax()

func ClampMin 

func ClampMin(tensor, minimum *Tensor) *Tensor

Clamps all elements in input to have minimum value of input tensor.

func ClampMin_ 

func ClampMin_(tensor, minimum *Tensor) *Tensor

In-place version of torch.ClampMin()

func Clamp_ 

func Clamp_(tensor, minimum, maximum *Tensor) *Tensor

In-place version of torch.Clamp()

func Concat 

func Concat(tensors []*Tensor, dim int64) *Tensor

Alias of torch.Cat.

func Concatenate 

func Concatenate(tensors []*Tensor, dim int64) *Tensor

Alias of torch.Cat.

func Decode 

func Decode(buffer []byte) (*Tensor, error)

Decode a pickled tensor back into a structured numerical format.

func Div 

func Div(tensor, other *Tensor) *Tensor

Divides input by other.

func Empty 

func Empty(size []int64, options *TensorOptions) *Tensor

Create a new tensor of given size filled with empty values.

func EmptyLike 

func EmptyLike(reference *Tensor) *Tensor

Create a tensor filled with empty values in the shape of a reference.

func Eq 

func Eq(tensor, other *Tensor) *Tensor

Create a new tensor comparing the element-wise equality of two tensors.

func Eye 

func Eye(n, m int64, options *TensorOptions) *Tensor

Create an NxM identity matrix.

func Flatten 

func Flatten(tensor *Tensor, startDim, endDim int64) *Tensor

Flattens input dimensions by reshaping them into a one-dimensional tensor. Only dimensions starting with start_dim and ending with end_dim are flattened. The order of elements in input is unchanged.

Unlike NumPy’s flatten, which always copies input’s data, this function may return the original object, a view, or copy. If no dimensions are flattened, then the original object input is returned. Otherwise, if input can be viewed as the flattened shape, then that view is returned. Finally, only if the input cannot be viewed as the flattened shape is input’s data copied. See (torch.Tensor).View() for details on when a view will be returned.

func Full 

func Full(size []int64, value float32, options *TensorOptions) *Tensor

Create a new tensor of given size filled with static values.

func FullLike 

func FullLike(reference *Tensor, value float32) *Tensor

Create a tensor filled with static values in the shape of a reference.

func Greater 

func Greater(tensor, other *Tensor) *Tensor

Compute tensor > other element-wise.

func GreaterEqual 

func GreaterEqual(tensor, other *Tensor) *Tensor

Compute tensor >= other element-wise.

func IndexSelect 

func IndexSelect(tensor *Tensor, dim int64, index *Tensor) *Tensor

Return a new tensor which indexes the input tensor along dimension dim using the entries in index which is a LongTensor. The returned tensor has the same number of dimensions as the original tensor (input). The dimth dimension has the same size as the length of index; other dimensions have the same size as in the original tensor.

func IsClose 

func IsClose(tensor, other *Tensor, rtol, atol float64) *Tensor

Create a new tensor describing the element-wise approximate equality of two tensors with given relative and absolute tolerances.

This function checks if all `tensor` and `other` satisfy the condition:

|tensor − other| <= atol + rtol * |other|

func IsFinite 

func IsFinite(tensor *Tensor) *Tensor

Test if each element of the tensor is finite or not.

func IsIn 

func IsIn(tensor, other *Tensor) *Tensor

Test if each element of tensor is in other. Returns a boolean tensor of the same shape as tensor that is True for elements in other and False otherwise.

func IsInf 

func IsInf(tensor *Tensor) *Tensor

Test if each element of the tensor is infinite (positive or negative) or not.

func IsNaN 

func IsNaN(tensor *Tensor) *Tensor

Returns a new tensor with boolean elements representing if each element of input is NaN or not. Complex values are considered NaN when either their real and/or imaginary part is NaN.

func IsNegInf 

func IsNegInf(tensor *Tensor) *Tensor

Test if each element of the tensor is negative infinity or not.

func IsPosInf 

func IsPosInf(tensor *Tensor) *Tensor

Test if each element of the tensor is positive infinity or not.

func IsReal 

func IsReal(tensor *Tensor) *Tensor

Return a new tensor with boolean elements representing if each element of input is real-valued or not. All real-valued types are considered real. Complex values are considered real when their imaginary part is 0.

func Less 

func Less(tensor, other *Tensor) *Tensor

Compute tensor < other element-wise.

func LessEqual 

func LessEqual(tensor, other *Tensor) *Tensor

Compute tensor <= other element-wise.

func Linspace 

func Linspace(begin, end float32, steps int64, options *TensorOptions) *Tensor

Create an linear space tensor from begin to end with given number of total steps.

func Load 

func Load(path string) (*Tensor, error)

Load a tensor from the given path.

func LogSoftmax 

func LogSoftmax(tensor *Tensor, dim int64) *Tensor

Apply a softmax followed by a logarithm.

While mathematically equivalent to log(softmax(x)), doing these two operations separately is slower and numerically unstable. This function uses an alternative formulation to compute the output and gradient correctly.

func LogicalAnd 

func LogicalAnd(tensor, other *Tensor) *Tensor

Compute the element-wise logical AND of the given input tensors. Zeros are treated as False and non-zeros are treated as True.

func LogicalNot 

func LogicalNot(tensor *Tensor) *Tensor

Computes the element-wise logical NOT of the given input tensor. If the input tensor is not a bool tensor, zeros are treated as False and non-zeros are treated as True.

func LogicalOr 

func LogicalOr(tensor, other *Tensor) *Tensor

Compute the element-wise logical OR of the given input tensors. Zeros are treated as False and non-zeros are treated as True.

func LogicalXor 

func LogicalXor(tensor, other *Tensor) *Tensor

Compute the element-wise logical XOR of the given input tensors. Zeros are treated as False and non-zeros are treated as True.

func Logspace 

func Logspace(begin, end float32, steps int64, base float64, options *TensorOptions) *Tensor

Create an logarithmic space tensor from begin to end with given number of total steps.

func MM 

func MM(a, b *Tensor) *Tensor

Perform a matrix multiplication of the matrices A and B. If A is an (n×m) tensor, B is an (m×p) tensor, the output will be an (n×p) tensor.

func Max 

func Max(tensor *Tensor) *Tensor

Reduce a tensor to its maximum value.

func Maximum 

func Maximum(tensor, other *Tensor) *Tensor

Computes the element-wise maximum of input and other.

func Mean 

func Mean(tensor *Tensor) *Tensor

Reduce a tensor to its mean value.

func MeanByDim 

func MeanByDim(tensor *Tensor, dim int, keep_dims bool) *Tensor

Reduce a tensor to its mean value along the given dimension.

func Median 

func Median(tensor *Tensor) *Tensor

Reduce a tensor to its median value.

func Min 

func Min(tensor *Tensor) *Tensor

Reduce a tensor to its minimum value.

func Minimum 

func Minimum(tensor, other *Tensor) *Tensor

Computes the element-wise minimum of input and other.

func Mul 

func Mul(tensor, other *Tensor) *Tensor

Multiplies input by other.

func NewTensor 

func NewTensor(data interface{}) *Tensor

Create a new tensor from a Go slice.

func NewTensorFromBlob 

func NewTensorFromBlob(data unsafe.Pointer, dtype Dtype, sizes []int64) (output *Tensor)

Create a new tensor that clones existing contiguous memory pointed to by data, of given data-type, and with given size. This function copies the input data, so subsequent in-place operations performed on the tensor will not mutate the input data.

func NotEqual 

func NotEqual(tensor, other *Tensor) *Tensor

Compute tensor != other element-wise.

func Ones 

func Ones(size []int64, options *TensorOptions) *Tensor

Create a new tensor of given size filled with ones.

func OnesLike 

func OnesLike(reference *Tensor) *Tensor

Create a tensor filled with ones in the shape of a reference.

func Permute 

func Permute(tensor *Tensor, dims ...int64) *Tensor

Return a view of the original tensor input with its dimensions permuted.

func Pow 

func Pow(tensor *Tensor, exponent float64) *Tensor

Take the power of each element in input with exponent and returns a tensor with the result.

func Rand 

func Rand(size []int64, options *TensorOptions) *Tensor

Create a new tensor of given size filled with uniform random values.

func RandInt 

func RandInt(size []int64, low int64, high int64, options *TensorOptions) *Tensor

Create a new tensor of given size filled with random integers in [low, high).

func RandIntLike 

func RandIntLike(reference *Tensor, low int64, high int64) *Tensor

Create a tensor filled with random integers in [low, high) in the shape of a reference.

func RandLike 

func RandLike(reference *Tensor) *Tensor

Create a tensor filled with uniform random values in the shape of a reference.

func RandN 

func RandN(size []int64, options *TensorOptions) *Tensor

Create a new tensor of given size filled with Gaussian random values.

func RandNLike 

func RandNLike(reference *Tensor) *Tensor

Create a tensor filled with Gaussian random values in the shape of a reference.

func Range 

func Range(begin, end, step float32, options *TensorOptions) *Tensor

Create an exclusive range tensor from begin position to end position along integer step size.

func Reshape 

func Reshape(tensor *Tensor, shape ...int64) *Tensor

Create a new tensor with an updated shape.

func Sigmoid 

func Sigmoid(tensor *Tensor) *Tensor

Computes the expit (also known as the logistic sigmoid function) of the elements of input.

func Slice 

func Slice(tensor *Tensor, dim, start, stop, step int64) *Tensor

Perform NumPy-like tensor slicing where dim is the dimension to slice along, start and stop determine the [start, stop) bounds of the index, and step describes the spacing between elements in the slice.

func Sqrt 

func Sqrt(tensor *Tensor) *Tensor

Take the square-root of input.

func Sqrt_ 

func Sqrt_(tensor *Tensor) *Tensor

In-place version of Sqrt().

func Square 

func Square(tensor *Tensor) *Tensor

Take the square of input.

func Square_ 

func Square_(tensor *Tensor) *Tensor

In-place version of Square().

func Squeeze 

func Squeeze(tensor *Tensor, dim ...int64) *Tensor

Return a tensor with all the dimensions of input of size 1 removed. For example, if input is of shape: (A×1×B×C×1×D) then the out tensor will be of shape: (A×B×C×D). When dim is given, a squeeze operation is done only in the given dimension. If input is of shape: (A×1×B), squeeze(input, 0) leaves the tensor unchanged, but squeeze(input, 1) will squeeze the tensor to the shape (A×B).

func Stack 

func Stack(tensors []*Tensor, dim int64) *Tensor

Concatenate a sequence of tensors along a new dimension. All tensors need to be of the same size.

func Std 

func Std(tensor *Tensor) *Tensor

Reduce a tensor to its standard deviation.

func StdByDim 

func StdByDim(tensor *Tensor, dim int, unbiased bool, keep_dims bool) *Tensor

Reduce a tensor to its standard deviation along the given dimension.

func Sub 

func Sub(tensor, other *Tensor, alpha float32) *Tensor

Subtracts other, scaled by alpha, from input.

func Sum 

func Sum(tensor *Tensor) *Tensor

Reduce a tensor to its sum.

func SumByDim 

func SumByDim(tensor *Tensor, dim int, keep_dims bool) *Tensor

Reduce a tensor to its sum along the given dimension.

func Tanh 

func Tanh(tensor *Tensor) *Tensor

Return a new tensor with the hyperbolic tangent of the elements of input.

func TensorFromBlob 

func TensorFromBlob(data unsafe.Pointer, dtype Dtype, sizes []int64) (output *Tensor)

Create a tensor view that wraps around existing contiguous memory pointed to by data, of given data-type, and with given size. This function does not copy the data buffer so in-place operations performed on the tensor will mutate the input buffer.

func Transpose 

func Transpose(tensor *Tensor, dim0, dim1 int64) *Tensor

Return a tensor that is a transposed version of input. The given dimensions dim0 and dim1 are swapped.

If input is a strided tensor then the resulting out tensor shares its underlying storage with the input tensor, so changing the content of one would change the content of the other.

If input is a sparse tensor then the resulting out tensor does not share the underlying storage with the input tensor.

If input is a sparse tensor with compressed layout (SparseCSR, SparseBSR, SparseCSC or SparseBSC) the arguments dim0 and dim1 must be both batch dimensions, or must both be sparse dimensions. The batch dimensions of a sparse tensor are the dimensions preceding the sparse dimensions.

func Unsqueeze 

func Unsqueeze(tensor *Tensor, dim int64) *Tensor

Return a new tensor with a dimension of size one inserted at the specified position. The returned tensor shares the same underlying data with this tensor. A dim value within the range [-input.dim() - 1, input.dim() + 1) can be used. Negative dim will correspond to unsqueeze() applied at dim = dim + input.dim() + 1.

func Var 

func Var(tensor *Tensor) *Tensor

Reduce a tensor to its variance.

func VarByDim 

func VarByDim(tensor *Tensor, dim int, unbiased, keep_dims bool) *Tensor

Reduce a tensor to its variance along the given dimension.

func Zeros 

func Zeros(size []int64, options *TensorOptions) *Tensor

Create a new tensor of given size filled with zeros.

func ZerosLike 

func ZerosLike(reference *Tensor) *Tensor

Create a tensor filled with zeros in the shape of a reference.

func (*Tensor) Abs 

func (tensor *Tensor) Abs() *Tensor

Take the absolute value of input.

func (*Tensor) Abs_ 

func (tensor *Tensor) Abs_() *Tensor

In-place version of Abs().

func (*Tensor) Add 

func (tensor *Tensor) Add(other *Tensor, alpha float32) *Tensor

Adds other, scaled by alpha, to input.

func (*Tensor) Add_ 

func (tensor *Tensor) Add_(other *Tensor, alpha float32) *Tensor

In-place version of Add().

func (*Tensor) All 

func (tensor *Tensor) All() *Tensor

Check if all values in the tensor evaluate to true.

func (*Tensor) AllByDim 

func (tensor *Tensor) AllByDim(dim int, keep_dims bool) *Tensor

Check if all values in the tensor evaluate to true along the given dimension.

func (*Tensor) AllClose 

func (tensor *Tensor) AllClose(other *Tensor, rtol, atol float64) bool

Return true if the tensor is approximately equal to another with given relative and absolute tolerances.

This function checks if all `tensor` and `other` satisfy the condition:

|tensor − other| <= atol + rtol * |other|

func (*Tensor) Any 

func (tensor *Tensor) Any() *Tensor

Check if any values in the tensor evaluate to true.

func (*Tensor) AnyByDim 

func (tensor *Tensor) AnyByDim(dim int, keep_dims bool) *Tensor

Check if any values in the tensor evaluate to true along the given dimension.

func (*Tensor) Argmax 

func (tensor *Tensor) Argmax() *Tensor

Reduce a tensor to its maximum index.

func (*Tensor) ArgmaxByDim 

func (tensor *Tensor) ArgmaxByDim(dim int, keep_dims bool) *Tensor

Reduce a tensor to its maximum index along the given dimension.

func (*Tensor) Argmin 

func (tensor *Tensor) Argmin() *Tensor

Reduce a tensor to its minimum index.

func (*Tensor) ArgminByDim 

func (tensor *Tensor) ArgminByDim(dim int, keep_dims bool) *Tensor

Reduce a tensor to its minimum index along the given dimension.

func (*Tensor) Backward 

func (tensor *Tensor) Backward()

Compute gradients based on the results of a forward pass through the tensor.

func (*Tensor) CastTo 

func (tensor *Tensor) CastTo(dtype Dtype) *Tensor

Create a new tensor with data cast to given type.

func (*Tensor) Clamp 

func (tensor *Tensor) Clamp(minimum, maximum *Tensor) *Tensor

Clamps all elements in input into the range [min, max].

func (*Tensor) ClampMax 

func (tensor *Tensor) ClampMax(maximum *Tensor) *Tensor

Clamps all elements in input to have maximum value of input tensor.

func (*Tensor) ClampMax_ 

func (tensor *Tensor) ClampMax_(maximum *Tensor) *Tensor

In-place version of torch.ClampMax()

func (*Tensor) ClampMin 

func (tensor *Tensor) ClampMin(minimum *Tensor) *Tensor

Clamps all elements in input to have minimum value of input tensor.

func (*Tensor) ClampMin_ 

func (tensor *Tensor) ClampMin_(minimum *Tensor) *Tensor

In-place version of torch.ClampMin()

func (*Tensor) Clamp_ 

func (tensor *Tensor) Clamp_(minimum, maximum *Tensor) *Tensor

In-place version of torch.Clamp()

func (*Tensor) Clone 

func (tensor *Tensor) Clone() (output *Tensor)

Create a clone of an existing tensor.

func (*Tensor) CopyTo 

func (tensor *Tensor) CopyTo(device *Device) *Tensor

Create a new tensor with data copied to given device.

func (*Tensor) Copy_ 

func (tensor *Tensor) Copy_(other *Tensor)

SetData sets the tensor data held by b to a

func (*Tensor) Detach 

func (tensor *Tensor) Detach() *Tensor

Create a new tensor with any taped gradients detached and grad disabled.

func (*Tensor) Dim 

func (tensor *Tensor) Dim() int64

Return the number of dimensions the tensor occupies.

func (*Tensor) Div 

func (tensor *Tensor) Div(other *Tensor) *Tensor

Divides input by other.

func (*Tensor) Div_ 

func (tensor *Tensor) Div_(other *Tensor) *Tensor

In-place version of Div().

func (*Tensor) Dtype 

func (tensor *Tensor) Dtype() Dtype

Return the data-type of the tensor.

func (*Tensor) Encode 

func (tensor *Tensor) Encode() ([]byte, error)

Encode a tensor into a pickled representation. Tensors are copied to the CPU before encoding.

func (*Tensor) Eq 

func (tensor *Tensor) Eq(other *Tensor) *Tensor

Create a new tensor comparing element-wise equality of the tensor to another.

func (*Tensor) Equal 

func (tensor *Tensor) Equal(other *Tensor) bool

Return true if the tensor is precisely equal to other element-wise.

func (*Tensor) Expand 

func (tensor *Tensor) Expand(shape ...int64) *Tensor

Expand the dimensions of the tensor.

func (*Tensor) ExpandAs 

func (tensor *Tensor) ExpandAs(other *Tensor) *Tensor

Expand the dimensions of the tensor to match that of a reference.

func (*Tensor) Flatten 

func (tensor *Tensor) Flatten(startDim, endDim int64) *Tensor

Flattens input dimensions by reshaping them into a one-dimensional tensor. Only dimensions starting with start_dim and ending with end_dim are flattened. The order of elements in input is unchanged.

Unlike NumPy’s flatten, which always copies input’s data, this function may return the original object, a view, or copy. If no dimensions are flattened, then the original object input is returned. Otherwise, if input can be viewed as the flattened shape, then that view is returned. Finally, only if the input cannot be viewed as the flattened shape is input’s data copied. See (torch.Tensor).View() for details on when a view will be returned.

func (*Tensor) Grad 

func (tensor *Tensor) Grad() *Tensor

Access the underlying gradients of the tensor.

func (*Tensor) Greater 

func (tensor *Tensor) Greater(other *Tensor) *Tensor

Compute tensor > other element-wise.

func (*Tensor) GreaterEqual 

func (tensor *Tensor) GreaterEqual(other *Tensor) *Tensor

Compute tensor >= other element-wise.

func (*Tensor) Index 

func (tensor *Tensor) Index(index *Tensor) *Tensor

Access elements in a tensor using an index tensor for reference.

func (*Tensor) IndexSelect 

func (tensor *Tensor) IndexSelect(dim int64, index *Tensor) *Tensor

Return a new tensor which indexes the input tensor along dimension dim using the entries in index which is a LongTensor. The returned tensor has the same number of dimensions as the original tensor (input). The dimth dimension has the same size as the length of index; other dimensions have the same size as in the original tensor.

func (*Tensor) IsClose 

func (tensor *Tensor) IsClose(other *Tensor, rtol, atol float64) *Tensor

Create a new tensor describing the element-wise approximate equality of the tensor to another with given relative and absolute tolerances.

This function checks if all `tensor` and `other` satisfy the condition:

|tensor − other| <= atol + rtol * |other|

func (*Tensor) IsComplex 

func (tensor *Tensor) IsComplex() bool

Return true if the tensor of a complex data-type.

func (*Tensor) IsConj 

func (tensor *Tensor) IsConj() bool

Return true if the (complex) tensor is in conjugated form.

func (*Tensor) IsFinite 

func (tensor *Tensor) IsFinite() *Tensor

Test if each element of the tensor is finite or not.

func (*Tensor) IsFloatingPoint 

func (tensor *Tensor) IsFloatingPoint() bool

Return true if the tensor of a floating-point data-type.

func (*Tensor) IsIn 

func (tensor *Tensor) IsIn(other *Tensor) *Tensor

Test if each element of tensor is in other. Returns a boolean tensor of the same shape as tensor that is True for elements in other and False otherwise.

func (*Tensor) IsInf 

func (tensor *Tensor) IsInf() *Tensor

Test if each element of the tensor is infinite (positive or negative) or not.

func (*Tensor) IsNaN 

func (tensor *Tensor) IsNaN() *Tensor

Returns a new tensor with boolean elements representing if each element of input is NaN or not. Complex values are considered NaN when either their real and/or imaginary part is NaN.

func (*Tensor) IsNegInf 

func (tensor *Tensor) IsNegInf() *Tensor

Test if each element of the tensor is negative infinity or not.

func (*Tensor) IsNonzero 

func (tensor *Tensor) IsNonzero() bool

Return true if the tensor is a single non-zero element (i.e., scalar.)

func (*Tensor) IsPosInf 

func (tensor *Tensor) IsPosInf() *Tensor

Test if each element of the tensor is positive infinity or not.

func (*Tensor) IsReal 

func (tensor *Tensor) IsReal() *Tensor

Return a new tensor with boolean elements representing if each element of input is real-valued or not. All real-valued types are considered real. Complex values are considered real when their imaginary part is 0.

func (*Tensor) Item 

func (tensor *Tensor) Item() interface{}

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

Users can do type assertion and get the value like:

``` 

if value, ok := a.Item().(float64); ok {
    // process the value
}

```

This function currently only supports signed data types.

func (*Tensor) Less 

func (tensor *Tensor) Less(other *Tensor) *Tensor

Compute tensor < other element-wise.

func (*Tensor) LessEqual 

func (tensor *Tensor) LessEqual(other *Tensor) *Tensor

Compute tensor <= other element-wise.

func (*Tensor) LogSoftmax 

func (tensor *Tensor) LogSoftmax(dim int64) *Tensor

Apply a softmax followed by a logarithm.

While mathematically equivalent to log(softmax(x)), doing these two operations separately is slower and numerically unstable. This function uses an alternative formulation to compute the output and gradient correctly.

func (*Tensor) LogicalAnd 

func (tensor *Tensor) LogicalAnd(other *Tensor) *Tensor

Compute the element-wise logical AND of the given input tensors. Zeros are treated as False and non-zeros are treated as True.

func (*Tensor) LogicalNot 

func (tensor *Tensor) LogicalNot() *Tensor

Computes the element-wise logical NOT of the given input tensor. If the input tensor is not a bool tensor, zeros are treated as False and non-zeros are treated as True.

func (*Tensor) LogicalOr 

func (tensor *Tensor) LogicalOr(other *Tensor) *Tensor

Compute the element-wise logical OR of the given input tensors. Zeros are treated as False and non-zeros are treated as True.

func (*Tensor) LogicalXor 

func (tensor *Tensor) LogicalXor(other *Tensor) *Tensor

Compute the element-wise logical XOR of the given input tensors. Zeros are treated as False and non-zeros are treated as True.

func (*Tensor) MM 

func (tensor *Tensor) MM(other *Tensor) *Tensor

Perform a matrix multiplication of the matrices A and B. If A is an (n×m) tensor, B is an (m×p) tensor, the output will be an (n×p) tensor.

func (*Tensor) Max 

func (tensor *Tensor) Max() *Tensor

Reduce a tensor to its maximum value.

func (*Tensor) MaxByDim 

func (tensor *Tensor) MaxByDim(dim int, keep_dims bool) ValueIndexPair

Reduce a tensor to its maximum value along the given dimension.

func (*Tensor) Maximum 

func (tensor *Tensor) Maximum(other *Tensor) *Tensor

Computes the element-wise maximum of input and other.

func (*Tensor) Mean 

func (tensor *Tensor) Mean() *Tensor

Reduce a tensor to its mean value.

func (*Tensor) MeanByDim 

func (tensor *Tensor) MeanByDim(dim int, keep_dims bool) *Tensor

Reduce a tensor to its mean value along the given dimension.

func (*Tensor) Median 

func (tensor *Tensor) Median() *Tensor

Reduce a tensor to its median value.

func (*Tensor) MedianByDim 

func (tensor *Tensor) MedianByDim(dim int, keep_dims bool) ValueIndexPair

Reduce a tensor to its median value along the given dimension.

func (*Tensor) Min 

func (tensor *Tensor) Min() *Tensor

Reduce a tensor to its minimum value.

func (*Tensor) MinByDim 

func (tensor *Tensor) MinByDim(dim int, keep_dims bool) ValueIndexPair

Reduce a tensor to its minimum value along the given dimension.

func (*Tensor) Minimum 

func (tensor *Tensor) Minimum(other *Tensor) *Tensor

Computes the element-wise minimum of input and other.

func (*Tensor) Mul 

func (tensor *Tensor) Mul(other *Tensor) *Tensor

Multiplies input by other.

func (*Tensor) Mul_ 

func (tensor *Tensor) Mul_(other *Tensor) *Tensor

In-place version of Mul().

func (*Tensor) NotEqual 

func (tensor *Tensor) NotEqual(other *Tensor) *Tensor

Compute tensor != other element-wise.

func (*Tensor) Numel 

func (tensor *Tensor) Numel() int64

Return the number of elements in the tensor.

func (*Tensor) Permute 

func (tensor *Tensor) Permute(dims ...int64) *Tensor

Return a view of the original tensor input with its dimensions permuted.

func (*Tensor) Pow 

func (tensor *Tensor) Pow(exponent float64) *Tensor

Take the power of each element in input with exponent and returns a tensor with the result.

func (*Tensor) RequiresGrad 

func (tensor *Tensor) RequiresGrad() bool

Return true if the tensor is taping gradients.

func (*Tensor) Reshape 

func (tensor *Tensor) Reshape(shape ...int64) *Tensor

Create a new tensor with an updated shape.

func (*Tensor) ReshapeAs 

func (tensor *Tensor) ReshapeAs(other *Tensor) *Tensor

Create a new tensor with an updated shape according to a reference tensor.

func (*Tensor) Save 

func (tensor *Tensor) Save(path string) error

Save the tensor to the given path.

func (*Tensor) SetData 

func (tensor *Tensor) SetData(other *Tensor)

Sets the tensor data to that of a separate reference tensor.

func (*Tensor) SetRequiresGrad 

func (tensor *Tensor) SetRequiresGrad(requiresGrad bool)

Set the gradient taping state of the tensor to a new value.

func (*Tensor) Shape 

func (tensor *Tensor) Shape() []int64

Return the shape of the tensor data.

func (*Tensor) Sigmoid 

func (tensor *Tensor) Sigmoid() *Tensor

Computes the expit (also known as the logistic sigmoid function) of the elements of input.

func (*Tensor) Slice 

func (tensor *Tensor) Slice(dim, start, stop, step int64) *Tensor

Perform NumPy-like tensor slicing where dim is the dimension to slice along, start and stop determine the [start, stop) bounds of the index, and step describes the spacing between elements in the slice.

func (*Tensor) Sort 

func (tensor *Tensor) Sort(dim int64, descending bool) ValueIndexPair

Sort the elements of the input tensor along a given dimension by value. If descending is True then the elements are sorted in descending order by value. A struct of (Values, Indices) is returned, where the values are the sorted values and indices are the indices of the elements in the original input tensor.

func (*Tensor) Sqrt 

func (tensor *Tensor) Sqrt() *Tensor

Take the square-root of input.

func (*Tensor) Sqrt_ 

func (tensor *Tensor) Sqrt_() *Tensor

In-place version of Sqrt().

func (*Tensor) Square 

func (tensor *Tensor) Square() *Tensor

Take the square of input.

func (*Tensor) Square_ 

func (tensor *Tensor) Square_() *Tensor

In-place version of Square().

func (*Tensor) Squeeze 

func (tensor *Tensor) Squeeze(dim ...int64) *Tensor

Return a tensor with all the dimensions of input of size 1 removed. For example, if input is of shape: (A×1×B×C×1×D) then the out tensor will be of shape: (A×B×C×D). When dim is given, a squeeze operation is done only in the given dimension. If input is of shape: (A×1×B), squeeze(input, 0) leaves the tensor unchanged, but squeeze(input, 1) will squeeze the tensor to the shape (A×B).

func (*Tensor) Std 

func (tensor *Tensor) Std() *Tensor

Reduce a tensor to its standard deviation.

func (*Tensor) StdByDim 

func (tensor *Tensor) StdByDim(dim int, unbiased bool, keep_dims bool) *Tensor

Reduce a tensor to its standard deviation along the given dimension.

func (*Tensor) StdMean 

func (tensor *Tensor) StdMean() (*Tensor, *Tensor)

Reduce a tensor to its mean value and standard deviation.

func (*Tensor) StdMeanByDim 

func (tensor *Tensor) StdMeanByDim(dim int, unbiased, keep_dims bool) (*Tensor, *Tensor)

Reduce a tensor to its mean value and standard deviation along given dimension.

func (*Tensor) String 

func (tensor *Tensor) String() string

Convert the tensor to a readable string representation.

func (*Tensor) Sub 

func (tensor *Tensor) Sub(other *Tensor, alpha float32) *Tensor

Subtracts other, scaled by alpha, from input.

func (*Tensor) Sub_ 

func (tensor *Tensor) Sub_(other *Tensor, alpha float32) *Tensor

In-place version of Sub().

func (*Tensor) Sum 

func (tensor *Tensor) Sum() *Tensor

Reduce a tensor to its sum.

func (*Tensor) SumByDim 

func (tensor *Tensor) SumByDim(dim int, keep_dims bool) *Tensor

Reduce a tensor to its sum along the given dimension.

func (*Tensor) Tanh 

func (tensor *Tensor) Tanh() *Tensor

Return a new tensor with the hyperbolic tangent of the elements of input.

func (*Tensor) To 

func (tensor *Tensor) To(device *Device, dtype Dtype) *Tensor

Create a new tensor with data copied to given device and cast to given type.

func (*Tensor) ToBytes 

func (tensor *Tensor) ToBytes() []byte

Convert a tensor to a raw binary representation as a byte slice. Note that this operation implies a safe copy of the underlying data. There are no garbage collector side-effects from usage of this function.

func (*Tensor) ToBytesUnsafe 

func (tensor *Tensor) ToBytesUnsafe() []byte

Convert a tensor to a raw binary representation as a byte slice. Note that this is a copy-free operation and simply returns a slice with its header updated to point to the underlying tensor data. If the tensor is garbage collected, the slice will be invalidated. Usage of this function typically requires calls to `runtime.KeepAlive(tensor)` for the input tensor to prevent accidentally freeing the tensor before the byte slice is consumed.

func (*Tensor) ToSlice 

func (tensor *Tensor) ToSlice() interface{}

Convert the Tensor to a Go slice. This function implies a flattening of the tensor to return a 1-dimensional vector.

func (*Tensor) TopK 

func (tensor *Tensor) TopK(k, dim int64, largest, sorted bool) ValueIndexPair

Return the k largest elements of the given input tensor along a given dimension. If largest is False then the k smallest elements are returned. A struct of (Values, Indices) is returned with the values and indices of the largest k elements of each row of the input tensor in the given dimension dim. The boolean option sorted if True, will make sure that the returned k elements are themselves sorted.

func (*Tensor) Transpose 

func (tensor *Tensor) Transpose(dim0, dim1 int64) *Tensor

Return a tensor that is a transposed version of input. The given dimensions dim0 and dim1 are swapped.

If input is a strided tensor then the resulting out tensor shares its underlying storage with the input tensor, so changing the content of one would change the content of the other.

If input is a sparse tensor then the resulting out tensor does not share the underlying storage with the input tensor.

If input is a sparse tensor with compressed layout (SparseCSR, SparseBSR, SparseCSC or SparseBSC) the arguments dim0 and dim1 must be both batch dimensions, or must both be sparse dimensions. The batch dimensions of a sparse tensor are the dimensions preceding the sparse dimensions.

func (*Tensor) Unsqueeze 

func (tensor *Tensor) Unsqueeze(dim int64) *Tensor

Return a new tensor with a dimension of size one inserted at the specified position. The returned tensor shares the same underlying data with this tensor. A dim value within the range [-input.dim() - 1, input.dim() + 1) can be used. Negative dim will correspond to unsqueeze() applied at dim = dim + input.dim() + 1.

func (*Tensor) Var 

func (tensor *Tensor) Var() *Tensor

Reduce a tensor to its variance.

func (*Tensor) VarByDim 

func (tensor *Tensor) VarByDim(dim int, unbiased, keep_dims bool) *Tensor

Reduce a tensor to its variance along the given dimension.

func (*Tensor) VarMean 

func (tensor *Tensor) VarMean() (*Tensor, *Tensor)

Reduce a tensor to its mean value and variance.

func (*Tensor) VarMeanByDim 

func (tensor *Tensor) VarMeanByDim(dim int, unbiased, keep_dims bool) (*Tensor, *Tensor)

Reduce a tensor to its mean value and variance along given dimension.

func (*Tensor) View 

func (tensor *Tensor) View(shape ...int64) *Tensor

Create a new tensor with an updated view of the underlying data.

func (*Tensor) ViewAs 

func (tensor *Tensor) ViewAs(other *Tensor) *Tensor

Create a new tensor with an updated view of the underlying data that matches that of a reference tensor.

type TensorOptions 

type TensorOptions struct {
	Pointer C.TensorOptions
}

TensorOptions wraps a C.TensorOptions.

func NewTensorOptions 

func NewTensorOptions() (options *TensorOptions)

Create a new TensorOptions.

func (*TensorOptions) Device 

func (options *TensorOptions) Device(device Device) *TensorOptions

Create a new TensorOptions with the given compute device.

func (*TensorOptions) Dtype 

func (options *TensorOptions) Dtype(value Dtype) *TensorOptions

Create a new TensorOptions with the given data type.

func (*TensorOptions) PinnedMemory 

func (options *TensorOptions) PinnedMemory(pinnedMemory bool) *TensorOptions

Create a new TensorOptions with the given memory pinning state.

func (*TensorOptions) RequiresGrad 

func (options *TensorOptions) RequiresGrad(requiresGrad bool) *TensorOptions

Create a new TensorOptions with the given gradient taping state.

type ValueIndexPair 

type ValueIndexPair struct {
	Values, Indices *Tensor
}

A representation of the return type for a paired value/index selection call.

func MaxByDim 

func MaxByDim(tensor *Tensor, dim int, keep_dims bool) ValueIndexPair

Reduce a tensor to its maximum value along the given dimension.

func MedianByDim 

func MedianByDim(tensor *Tensor, dim int, keep_dims bool) ValueIndexPair

Reduce a tensor to its median value along the given dimension.

func MinByDim 

func MinByDim(tensor *Tensor, dim int, keep_dims bool) ValueIndexPair

Reduce a tensor to its minimum value along the given dimension.

func Sort 

func Sort(tensor *Tensor, dim int64, descending bool) ValueIndexPair

Sort the elements of the input tensor along a given dimension by value. If descending is True then the elements are sorted in descending order by value. A struct of (Values, Indices) is returned, where the values are the sorted values and indices are the indices of the elements in the original input tensor.

func TopK 

func TopK(tensor *Tensor, k, dim int64, largest, sorted bool) ValueIndexPair

Return the k largest elements of the given input tensor along a given dimension. If largest is False then the k smallest elements are returned. A struct of (Values, Indices) is returned with the values and indices of the largest k elements of each row of the input tensor in the given dimension dim. The boolean option sorted if True, will make sure that the returned k elements are themselves sorted.

Source Files

View all Source files

Directories

Path Synopsis
cmd
faster_rcnn_inference
imagenet_inference
nn
functional
initialize
vision
ops
transforms
transforms/functional

Jump to

Keyboard shortcuts

? : This menu
/ : Search site
f or F : Jump to
y or Y : Canonical URL
go.dev uses cookies from Google to deliver and enhance the quality of its services and to analyze traffic. Learn more.