Trait dfdx::optim::Optimizer

pub trait Optimizer<M: CanUpdateWithGradients> {
    fn update(
        &mut self,
        module: &mut M,
        gradients: Gradients
    ) -> Result<(), UnusedParamsError>;
}

All optimizers must implement the update function, which takes an object that implements CanUpdateWithGradients, and calls CanUpdateWithGradients::update.

Notes

1. CanUpdateWithGradients requires an object that implements crate::gradients::GradientProvider. A common implementation involves implementing both Optimizer and crate::gradients::GradientProvider on one struct, and passing self to CanUpdateWithGradients::update. See super::Sgd for an example of implementing this trait.

2. Update takes ownership of Gradients, so update cannot be called with the same gradients object.

3. Optimizer itself is generic over M, not the update method. This means a single optimizer object can only work on objects of type M. This also requires you to specify the model up front for the optimizer.

Required Methods

fn update(
    &mut self,
    module: &mut M,
    gradients: Gradients
) -> Result<(), UnusedParamsError>

Updates all of module’s parameters using gradients.

Requires a &mut self because the optimizer may change some internally tracked values.

Implementors

impl<M: CanUpdateWithGradients> Optimizer<M> for Adam<M>

impl<M: CanUpdateWithGradients> Optimizer<M> for RMSprop<M>

impl<M: CanUpdateWithGradients> Optimizer<M> for Sgd<M>