Function l1_loss_with_reduction 

pub fn l1_loss_with_reduction<B>(
    prediction: &Tensor<B>,
    target: &Tensor<B>,
    reduction: Reduction,
) -> Result<Tensor<B>, Error>
where
    B: Backend,

L1 Loss with configurable reduction.