Struct ReduceLROnPlateauScheduler 

pub struct ReduceLROnPlateauScheduler {
    pub factor: f64,
    pub patience: usize,
    pub min_lr: f64,
    pub threshold: f64,
    pub cooldown: usize,
    /* private fields */
}

Reduce learning rate on plateau (metric-based adaptive scheduler).

Reduces learning rate when a metric (e.g., validation loss) has stopped improving. This scheduler requires explicit metric updates via step_with_metric().

Fields

factor: f64

Decay factor.

patience: usize

Number of epochs with no improvement after which LR will be reduced.

min_lr: f64

Minimum LR.

threshold: f64

Threshold for measuring improvement (relative).

cooldown: usize

Number of epochs to wait before resuming normal operation after LR reduction.

Implementations

impl ReduceLROnPlateauScheduler

pub fn new( initial_lr: f64, mode: PlateauMode, factor: f64, patience: usize, threshold: f64, min_lr: f64, cooldown: usize, ) -> Self

Create a new ReduceLROnPlateau scheduler.

Arguments

• initial_lr - Initial learning rate
• mode - Whether to minimize or maximize the metric
• factor - Factor by which to reduce LR (new_lr = lr * factor)
• patience - Number of epochs with no improvement to wait
• threshold - Threshold for measuring improvement (relative)
• min_lr - Minimum LR (won’t reduce below this)
• cooldown - Cooldown epochs after LR reduction

pub fn step_with_metric(&mut self, optimizer: &mut dyn Optimizer, metric: f64)

Step with a metric value.

This should be called with the validation metric at the end of each epoch.

Trait Implementations

impl Clone for ReduceLROnPlateauScheduler

fn clone(&self) -> ReduceLROnPlateauScheduler

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for ReduceLROnPlateauScheduler

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

Formats the value using the given formatter.

impl LrScheduler for ReduceLROnPlateauScheduler

fn step(&mut self, _optimizer: &mut dyn Optimizer)

Update learning rate based on current step/epoch.

fn get_lr(&self) -> f64

Get current learning rate.

fn state_dict(&self) -> HashMap<String, f64>

Get scheduler state as a dictionary.

fn load_state_dict(&mut self, state: &HashMap<String, f64>) -> TrainResult<()>

Load scheduler state from a dictionary.