Struct GroupNorm 

pub struct GroupNorm { /* private fields */ }

Group Normalization (Wu & He, 2018).

Divides channels into groups and normalizes within each group. Unlike BatchNorm, GroupNorm is independent of batch size, making it suitable for small batch training (e.g., object detection, segmentation).

Shape

• Input: (N, C, *) where C must be divisible by num_groups
• Output: Same as input

Example

use aprender::nn::{GroupNorm, Module};
use aprender::autograd::Tensor;

let norm = GroupNorm::new(32, 256);  // 32 groups, 256 channels
let x = Tensor::randn(&[4, 256, 14, 14]);
let y = norm.forward(&x);

Implementations

impl GroupNorm

pub fn new(num_groups: usize, num_channels: usize) -> Self

Create a new GroupNorm layer.

Arguments

• num_groups - Number of groups to divide channels into
• num_channels - Number of channels (must be divisible by num_groups)

Panics

Panics if num_channels is not divisible by num_groups.

pub fn with_eps(num_groups: usize, num_channels: usize, eps: f32) -> Self

Create GroupNorm with custom epsilon.

pub fn without_affine(num_groups: usize, num_channels: usize) -> Self

Create GroupNorm without learnable parameters.

pub fn num_groups(&self) -> usize

Get number of groups.

pub fn num_channels(&self) -> usize

Get number of channels.

Trait Implementations

impl Debug for GroupNorm

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

Formats the value using the given formatter.

impl Module for GroupNorm

fn forward(&self, input: &Tensor) -> Tensor

Perform forward computation.

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

Get references to all learnable parameters.

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

Get mutable references to all learnable parameters.

fn refresh_caches(&mut self)

Refresh any cached computations after parameters have been modified.

fn train(&mut self)

Set the module to training mode.

fn eval(&mut self)

Set the module to evaluation mode.

fn training(&self) -> bool

Check if the module is in training mode.

fn zero_grad(&mut self)

Zero out gradients for all parameters.

fn num_parameters(&self) -> usize

Get the number of learnable parameters.