Trait TlAutodiff 

pub trait TlAutodiff: TlExecutor {
    type Tape;

    // Required methods
    fn forward(
        &mut self,
        graph: &EinsumGraph,
    ) -> Result<Self::Tensor, Self::Error>;
    fn backward(
        &mut self,
        graph: &EinsumGraph,
        loss: &Self::Tensor,
    ) -> Result<Self::Tape, Self::Error>;
}

Automatic differentiation interface.

Extends TlExecutor with forward/backward pass capabilities for training.

Required Associated Types

type Tape

Required Methods

fn forward(&mut self, graph: &EinsumGraph) -> Result<Self::Tensor, Self::Error>

Execute forward pass on an EinsumGraph.

fn backward( &mut self, graph: &EinsumGraph, loss: &Self::Tensor, ) -> Result<Self::Tape, Self::Error>

Execute backward pass to compute gradients.

Implementors

impl TlAutodiff for DummyExecutor

type Tape = HashMap<usize, DummyTensor>