Struct BatchNorm 

pub struct BatchNorm { /* private fields */ }

Batch normalization operation

Implementations

impl BatchNorm

pub fn new(epsilon: f32) -> Self

Create a new batch normalization layer

pub fn forward( &self, input: &[f32], mean: &[f32], variance: &[f32], gamma: &[f32], beta: &[f32], output: &mut [f32], batch_size: usize, features: usize, )

Apply batch normalization to a batch of data Input shape: [batch_size, features] mean, variance, gamma, beta shape: [features]

pub fn forward_f16( &self, input: &[F16], mean: &[F16], variance: &[F16], gamma: &[F16], beta: &[F16], output: &mut [F16], batch_size: usize, features: usize, )

Apply batch normalization with FP16 precision

pub fn compute_stats( input: &[f32], mean: &mut [f32], variance: &mut [f32], batch_size: usize, features: usize, )

Compute batch statistics (mean and variance)