Struct GradStore 

pub struct GradStore<B: Backend> { /* private fields */ }

Stores gradients for all tensors in a computation graph.

After calling tensor.backward(), you receive a GradStore. Use grads.get(&tensor) to retrieve the gradient for any tensor.

Implementations

impl<B: Backend> GradStore<B>

pub fn new() -> Self

Create a new empty GradStore.

pub fn get(&self, tensor: &Tensor<B>) -> Option<&Tensor<B>>

Get the gradient of a tensor (if it exists).

pub fn accumulate(&mut self, id: TensorId, grad: Tensor<B>) -> Result<()>

Accumulate gradient for a tensor. If a gradient already exists for this tensor, add the new one to it. This handles the case where a tensor is used in multiple operations.

Trait Implementations

impl<B: Backend> Clone for GradStore<B>

fn clone(&self) -> Self

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl<B: Backend> Default for GradStore<B>

fn default() -> Self

Returns the “default value” for a type.