Struct ConvConfig 

pub struct ConvConfig {
    pub kernel_size: Vec<usize>,
    pub stride: Vec<usize>,
    pub padding: Vec<usize>,
    pub dilation: Vec<usize>,
    pub groups: usize,
}

Convolution configuration specifying kernel size, stride, padding, dilation, and groups.

Fields

kernel_size: Vec<usize>

Kernel size for each spatial dimension (e.g. [3, 3]).

stride: Vec<usize>

Stride for each spatial dimension (e.g. [1, 1]).

padding: Vec<usize>

Zero-padding on each side for each spatial dimension (e.g. [1, 1]).

dilation: Vec<usize>

Dilation factor for each spatial dimension (e.g. [1, 1]).

groups: usize

Number of groups for grouped convolution (1 = standard convolution).

Implementations

impl ConvConfig

pub fn new(kernel_size: Vec<usize>) -> Self

Create a new convolution config with the given kernel size. Defaults: stride=1, padding=0, dilation=1, groups=1.

pub fn with_stride(self, stride: Vec<usize>) -> Self

Set the stride (builder pattern).

pub fn with_padding(self, padding: Vec<usize>) -> Self

Set the padding (builder pattern).

pub fn with_dilation(self, dilation: Vec<usize>) -> Self

Set the dilation (builder pattern).

pub fn with_groups(self, groups: usize) -> Self

Set the number of groups (builder pattern).

pub fn output_size(&self, input_size: usize, dim: usize) -> usize

Compute the output size for one spatial dimension.

Formula: (input_size + 2*padding - dilation*(kernel-1) - 1) / stride + 1

pub fn validate(&self) -> Result<(), ConvError>

Validate the configuration, returning an error if any parameter is invalid.

pub fn num_spatial_dims(&self) -> usize

Number of spatial dimensions (length of kernel_size).

Trait Implementations

impl Clone for ConvConfig

fn clone(&self) -> ConvConfig

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for ConvConfig

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

Formats the value using the given formatter.