Struct MaxPool2D 

pub struct MaxPool2D { /* private fields */ }

Implementations

impl MaxPool2D

pub fn new( kernel_size: (usize, usize), stride: Option<(usize, usize)>, ) -> MaxPool2D

Trait Implementations

impl Clone for MaxPool2D

fn clone(&self) -> MaxPool2D

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl<T> Layer<T> for MaxPool2D

where T: Clone + Default + Zero + PartialOrd + Send + Sync + 'static + Pod + Zeroable,

fn forward(&self, input: &Tensor<T>) -> Result<Tensor<T>, TensorError>

fn parameters(&self) -> Vec<&Tensor<T>>

fn parameters_mut(&mut self) -> Vec<&mut Tensor<T>>

fn set_training(&mut self, _training: bool)

fn clone_box(&self) -> Box<dyn Layer<T>>

fn forward_multi(&self, inputs: &[&Tensor<T>]) -> Result<Tensor<T>, TensorError>

Forward pass for layers that accept multiple input tensors (e.g. merge/combine layers). The default implementation only supports the single-input case: it requires exactly one input and delegates to Layer::forward, so existing single-input layer implementations are completely unaffected by this method’s addition (no breaking change). Layers that genuinely need multiple inputs should override this method.

fn layer_type(&self) -> LayerType

Returns the type of this layer for ONNX export and introspection

fn set_weight(&mut self, _weight: Tensor<T>) -> Result<(), TensorError>

Set weight tensor for layers that support weights Default implementation returns an error

fn set_bias(&mut self, _bias: Option<Tensor<T>>) -> Result<(), TensorError>

Set bias tensor for layers that support bias Default implementation returns an error