Struct BatchNorm

pub struct BatchNorm<B: Backend> {
    pub eps: f32,
    pub momentum: f32,
    pub track_running_stats: bool,
    pub training: bool,
    pub weight: Option<Tensor<B>>,
    pub bias: Option<Tensor<B>>,
    pub running_mean: Tensor<B>,
    pub running_var: Tensor<B>,
    pub num_batches_tracked: Tensor<B>,
}

Batch norm

By default this module has learnable affine parameters, set weight and bias to None to remove them.

Fields

eps: f32

a value added to the denominator for numerical stability. Default: 1e-5

momentum: f32

the value used for the running_mean and running_var computation. Can be set to None for cumulative moving average (i.e. simple average). Default: 0.1

track_running_stats: bool

When set to True, this module tracks the running mean and variance, and when set to False, this module does not track such statistics, and initializes statistics buffers running_mean and running_var as None. When these buffers are None, this module always uses batch statistics

training: bool

Is it training or inference? (for running mean and var)

weight: Option<Tensor<B>>

weight

bias: Option<Tensor<B>>

bias

running_mean: Tensor<B>

weight

running_var: Tensor<B>

bias

num_batches_tracked: Tensor<B>

Number of tracked batches

Implementations

impl<B: Backend> BatchNorm<B>

pub fn forward(&mut self, x: &Tensor<B>) -> Tensor<B>

Forward function for layer_norm.

Trait Implementations

impl<'a, B: Backend> IntoIterator for &'a BatchNorm<B>

type Item = &'a Tensor<B>

The type of the elements being iterated over.

type IntoIter = IntoIter<&'a Tensor<B>>

Which kind of iterator are we turning this into?

fn into_iter(self) -> Self::IntoIter

Creates an iterator from a value.

impl<'a, B: Backend> IntoIterator for &'a mut BatchNorm<B>

type Item = &'a mut Tensor<B>

The type of the elements being iterated over.

type IntoIter = IntoIter<<&'a mut BatchNorm<B> as IntoIterator>::Item>

Which kind of iterator are we turning this into?

fn into_iter(self) -> Self::IntoIter

Creates an iterator from a value.