Function batch_norm 

pub fn batch_norm(
    input: &Array,
    running_mean: &mut Array,
    running_var: &mut Array,
    weight: &Array,
    bias: &Array,
    training: bool,
    momentum: f32,
    eps: f32,
) -> Result<Array>

Batch Normalization 1D

Applies Batch Normalization over a 3D input (Batch, Channels, Length).

Arguments

• input - Input tensor [B, C, L]
• running_mean - Running mean stats [C] (In-place update during training)
• running_var - Running variance stats [C] (In-place update during training)
• weight - Learnable gamma [C]
• bias - Learnable beta [C]
• training - If true, uses batch stats and updates running stats. If false, uses running stats.
• momentum - Momentum for running stats check (default use 0.1)
• eps - Epsilon for stability