Struct QuantizedMaxPool2d 

pub struct QuantizedMaxPool2d { /* private fields */ }

Quantized Max Pooling 2D

Operates directly in INT8 domain without scale changes. Output has the same scale and zero-point as input.

Implementations

impl QuantizedMaxPool2d

pub fn new(kernel_size: usize, stride: usize, padding: usize) -> Self

Create a new quantized max pooling layer

pub fn forward_int8( &self, input: &[u8], input_shape: &[usize], scale: f32, zero_point: u8, ) -> CnnResult<(Vec<u8>, Vec<usize>, f32, u8)>

Forward pass with INT8 input

Arguments

• input - Quantized u8 input tensor (NHWC layout)
• input_shape - Input shape [N, H, W, C]
• scale - Input/output scale (unchanged)
• zero_point - Input/output zero point (unchanged)

Trait Implementations

impl Clone for QuantizedMaxPool2d

fn clone(&self) -> QuantizedMaxPool2d

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for QuantizedMaxPool2d

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

Formats the value using the given formatter.