Function hard_shrink 

pub fn hard_shrink<const D: usize>(tensor: Tensor<D>, lambda: f64) -> Tensor<D>

Applies the HardShrink function element-wise.

hard_shrink(x) = x if x > lambda, x if x < -lambda, 0 otherwise

Arguments

• lambda: the lambda value for the Hard Shrink formulation. Default is 0.5.