Struct ReduceDescriptor 

pub struct ReduceDescriptor<const N: usize> {
    pub kind: ReduceKind,
    pub input_shape: [i32; N],
    pub reduce_axis: u8,
    pub element: ElementKind,
    pub correction: i32,
}

Descriptor for a single-axis reduction.

input_shape is the shape of the input tensor. reduce_axis is the axis to reduce (0 ≤ reduce_axis < rank). Output shape is the input shape with [reduce_axis] collapsed to size 1 (keepdim convention — caller squeezes if they want).

Fields

kind: ReduceKind

Which reduction to apply (Sum / Mean / Max / …).

input_shape: [i32; N]

Input tensor shape.

reduce_axis: u8

Axis to reduce along. Must satisfy 0 <= reduce_axis < N.

element: ElementKind

Element type.

correction: i32

Bessel correction for Var / Std only. 1 = sample variance (PyTorch default), 0 = population variance. Ignored by other reductions.

Implementations

impl<const N: usize> ReduceDescriptor<N>

pub fn output_shape(&self) -> [i32; N]

Compute the output shape (input shape with reduce axis = 1).

Trait Implementations

impl<const N: usize> Clone for ReduceDescriptor<N>

fn clone(&self) -> ReduceDescriptor<N>

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl<const N: usize> Copy for ReduceDescriptor<N>

impl<const N: usize> Debug for ReduceDescriptor<N>

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.