Function dfdx::tensor_ops::normalize

source ·

pub fn normalize<Ax: Axes, S: Shape + ReduceShape<Ax>, E: Dtype, D: Device<E>, T: Tape<E, D>>(
    t: Tensor<S, E, D, T>,
    epsilon: impl Into<f64>
) -> Tensor<S, E, D, T>

Expand description

Normalizes t to have mean 0.0 and stddev 1.0 along Ax. epsilon is used during stddev. Computes (t - t.mean(Ax)) / t.std(Ax, epsilon).

Normalizing a single axis:

let t: Tensor<Rank2<2, 3>, f32, _> = dev.zeros();
let _ = t.normalize::<Axis<1>>(1e-5);