Module layout 

Structs

BlockScaledLayout

BlockScaledLayoutCompilationArg

BlockScaledLayoutExpand

BlockScaledLayoutLaunch

PerTensorLayout

PerTensorLayoutCompilationArg

PerTensorLayoutExpand

PerTensorLayoutLaunch

Enums

ScalesLayout

Layout for quantization scales, indexed by quant element index and returns the corresponding scale based on the quantization type.

ScalesLayoutArgs

ScalesLayoutCompilationArg

ScalesLayoutExpand

Functions

scales_layout

scales_view

Create a scales view from the values and scales handle, line size and quantization scheme. values should be the quantized tensor, and will be adjusted by num_quants.

Type Aliases

ScalesView

[TensorView] with a linear layout inferred from the shape/strides at launch. Useful for elementwise kernels.

ScalesViewLaunch

Launch type for [LinearTensorView].