Trait ForwardRule 

pub trait ForwardRule<V>
where
    V: Differentiable,
{
    // Required method
    fn pushforward(
        &self,
        tangents: &[Option<&<V as Differentiable>::Tangent>],
    ) -> Result<<V as Differentiable>::Tangent, AutodiffError>;
}

Forward-mode AD rule interface (frule).

Named after Julia’s ChainRules.jl convention: frule computes pushforward.

The type parameter V is the differentiable value type (e.g., Tensor<f64>).

Examples

use chainrules_core::{ForwardRule, Differentiable, AdResult};

struct MyFrule;
impl<V: Differentiable> ForwardRule<V> for MyFrule {
    fn pushforward(&self, tangents: &[Option<&V::Tangent>])
        -> AdResult<V::Tangent> {
        todo!()
    }
}

Required Methods

fn pushforward( &self, tangents: &[Option<&<V as Differentiable>::Tangent>], ) -> Result<<V as Differentiable>::Tangent, AutodiffError>

Computes output tangent from input tangents (pushforward).

Implementors