Struct QHM 

pub struct QHM { /* private fields */ }

Quasi-Hyperbolic Momentum optimizer.

QHM interpolates between the current gradient and the momentum buffer, providing a generalization of both momentum and Nesterov acceleration. Update rule: p = p - lr * (nu * g + (1 - nu) * momentum)

Implementations

impl QHM

pub fn new(config: QHMConfig) -> Self

Create a new QHM optimizer.

pub fn with_defaults(learning_rate: f32, momentum: f32, nu: f32) -> Self

Create QHM with default configuration.

pub fn get_config(&self) -> &QHMConfig

Get the configuration.

pub fn set_config(&mut self, config: QHMConfig)

Update configuration.

Trait Implementations

impl Debug for QHM

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

Formats the value using the given formatter.

impl OptimizerState for QHM

fn zero_grad(&mut self) -> Result<()>

Zero out gradients

fn step(&mut self, parameters: &mut [Tensor]) -> Result<()>

Perform optimization step

fn get_lr(&self) -> f32

Get current learning rate

fn set_lr(&mut self, lr: f32)

Set learning rate

fn state_dict(&self) -> Result<HashMap<String, Tensor>>

Save optimizer state to dictionary

fn load_state_dict(&mut self, state: HashMap<String, Tensor>) -> Result<()>

Load optimizer state from dictionary