Struct BatchNorm2d 

pub struct BatchNorm2d<B>
where
    B: Backend,
{ /* private fields */ }

2D Batch Normalization layer for convolutional networks.

Normalizes each channel across the batch: for input [N, C, H, W], mean and variance are computed over (N, H, W) for each of C channels.

Examples

let bn = BatchNorm2d::<CpuBackend>::new(16, 1e-5, 0.1, DType::F64, &dev)?;
let x: [batch, 16, H, W] tensor
let y = bn.forward(&x)?; // normalized, same shape

Implementations

impl<B> BatchNorm2d<B>

where B: Backend,

pub fn new( num_features: usize, eps: f64, momentum: f64, dtype: DType, device: &<B as Backend>::Device, ) -> Result<BatchNorm2d<B>, Error>

Create a new BatchNorm2d layer.

Arguments

• num_features: number of channels (C)
• eps: numerical stability constant (typically 1e-5)
• momentum: EMA momentum for running stats (typically 0.1)
• dtype: data type for learnable parameters
• device: device

pub fn train(&self)

Set training mode (use batch statistics).

pub fn eval(&self)

Set evaluation mode (use running statistics).

pub fn is_training(&self) -> bool

Whether the module is in training mode.

pub fn num_features(&self) -> usize

pub fn eps(&self) -> f64

pub fn weight(&self) -> &Tensor<B>

pub fn bias(&self) -> &Tensor<B>

pub fn from_tensors( weight: Tensor<B>, bias: Tensor<B>, eps: f64, ) -> Result<BatchNorm2d<B>, Error>

Create from existing weight and bias tensors (for executor/model loading). Initializes running stats to mean=0, var=1.

Trait Implementations

impl<B> Module<B> for BatchNorm2d<B>

where B: Backend,

fn forward(&self, x: &Tensor<B>) -> Result<Tensor<B>, Error>

Forward pass: batch-normalize each channel.

Training: use batch mean/var, update running stats. Eval: use running mean/var.

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

Return all trainable parameters of this module. The optimizer uses these to update weights during training.

fn named_parameters(&self) -> Vec<(String, Tensor<B>)>

Return all trainable parameters with human-readable names.

fn set_training(&self, training: bool)

Set training or evaluation mode.

fn is_training(&self) -> bool

Whether the module is in training mode (default: true).

fn train(&self)

Convenience: set training mode.

fn eval(&self)

Convenience: set evaluation mode.

fn num_parameters(&self) -> usize

Total number of scalar parameters in this module.

fn trainable_params_count(&self) -> usize

Number of trainable (variable) parameters.

fn frozen_parameters(&self) -> Vec<Tensor<B>>

Freeze all parameters: returns new parameter tensors with is_variable = false, preventing gradient accumulation.

fn state_dict(&self) -> Vec<(String, Tensor<B>)>

Returns a state_dict-style map of parameter name → tensor.