Struct SGD 

pub struct SGD { /* private fields */ }

Stochastic Gradient Descent optimizer with momentum.

Update rule:

v_t = momentum * v_{t-1} + grad
param = param - lr * v_t

With Nesterov momentum:

v_t = momentum * v_{t-1} + grad
param = param - lr * (momentum * v_t + grad)

Implementations

impl SGD

pub fn new(params: Vec<&mut Tensor>, lr: f32) -> Self

Create a new SGD optimizer.

Arguments

• params - Mutable references to parameter tensors
• lr - Learning rate

pub fn with_momentum(params: Vec<&mut Tensor>, lr: f32, momentum: f32) -> Self

Create SGD with momentum.

pub fn nesterov(self) -> Self

Enable Nesterov momentum.

pub fn weight_decay(self, wd: f32) -> Self

Set weight decay (L2 regularization).

impl SGD

pub fn step_with_params(&mut self, params: &mut [&mut Tensor])

Perform optimization step with direct tensor access.

This is the recommended way to use SGD in a training loop.

Trait Implementations

impl Debug for SGD

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Optimizer for SGD

fn step(&mut self)

Perform a single optimization step using computed gradients.

fn zero_grad(&mut self)

Zero all parameter gradients.

fn lr(&self) -> f32

Get current learning rate.

fn set_lr(&mut self, lr: f32)

Set learning rate (for schedulers).