Struct SGD 

pub struct SGD<B>
where
    B: Backend,
{ /* private fields */ }

Stochastic Gradient Descent optimizer with optional momentum.

Parameters

• lr: learning rate (typical: 0.01 - 0.1)
• momentum: momentum factor (typical: 0.9, 0 = no momentum)
• weight_decay: L2 regularization coefficient (typical: 1e-4)

Implementations

impl<B> SGD<B>

where B: Backend,

pub fn new( params: Vec<Tensor<B>>, lr: f64, momentum: f64, weight_decay: f64, ) -> SGD<B>

Create a new SGD optimizer.

Arguments

• params: the model parameters to optimize
• lr: learning rate
• momentum: momentum factor (0 for vanilla SGD)
• weight_decay: L2 regularization strength

pub fn update_params(&mut self, new_params: Vec<Tensor<B>>)

Update the parameter references (needed after step() returns new tensors).

pub fn params(&self) -> &[Tensor<B>]

Access current parameters.

pub fn params_mut(&mut self) -> &mut Vec<Tensor<B>>

Mutable access to current parameters (for checkpoint loading).

Trait Implementations

impl<B> Optimizer<B> for SGD<B>

where B: Backend,

fn step(&mut self, grads: &GradStore<B>) -> Result<Vec<Tensor<B>>, Error>

Perform one optimization step.

fn learning_rate(&self) -> f64

Return the current learning rate.

fn set_learning_rate(&mut self, lr: f64)

Set a new learning rate (for learning rate scheduling).

impl<B> Stateful for SGD<B>

where B: Backend,

fn state_dict(&self) -> OptimizerState

Export the optimizer’s internal state as a serializable dictionary.

fn load_state_dict(&mut self, state: &OptimizerState) -> Result<(), Error>

Restore the optimizer’s internal state from a previously saved dictionary.