Module ops 

Operations on tensors module.

Modules

attention

Module with attention operations.

conv

Module with convolution operations.

grid_sample

Module for grid_sample operations

pool

Module with pooling operations.

unfold

Module with unfold operations.

Structs

Conv2dBackward

Gradient computed during the backward pass for each tensor used by conv2d.

Conv3dBackward

Gradient computed during the backward pass for each tensor used by conv3d.

ConvOptions

Convolution options.

ConvTransposeOptions

Transposed convolution options.

DeformConv2dBackward

Gradient computed during the backward pass for each tensor used by deform_conv2d.

DeformConvOptions

Convolution options.

GridSampleOptions

Options for grid sampling operations.

InterpolateBackward

Gradient computed during the backward pass for each tensor used by interpolate.

InterpolateOptions

Interpolation options.

MaxPool1dBackward

Gradient computed during the backward pass for each tensor used by max_pool1d.

MaxPool1dWithIndices

Results from max_pool1d.

MaxPool2dBackward

Gradient computed during the backward pass for each tensor used by max_pool2d.

MaxPool2dWithIndices

Results from max_pool2d.

TransactionPrimitive

Contains all tensor primitives that are going to be read.

TransactionPrimitiveData

Contains all data related to a transaction.

UnfoldOptions

Unfold operation options.

Enums

GridSamplePaddingMode

Padding mode for grid sampling when coordinates are out of bounds.

InterpolateMode

Algorithm used for upsampling.

PadMode

Padding mode for tensor pad operations.

Traits

ActivationOps

Activation function operations.

BoolTensorOps

Bool Tensor API for basic operations, see Tensor for documentation on each function.

FloatTensorOps

Operations on float tensors.

IntTensorOps

Int Tensor API for basic and numeric operations, see Tensor for documentation on each function.

ModuleOps

Module operations trait.

QTensorOps

Operations on quantized tensors.

TransactionOps

Operations that are sync by nature and that can be batch together in transactions to improve compute utilization with efficient laziness.

Type Aliases

BoolElem

Boolean element type used by backend.

BoolTensor

Boolean tensor primitive type used by the backend.

Device

Device type used by the backend.

FloatElem

Float element type used by backend.

FloatTensor

Float tensor primitive type used by the backend.

IntElem

Integer element type used by backend.

IntTensor

Integer tensor primitive type used by the backend.

QuantizedTensor

Quantized tensor primitive type used by the backend.