Struct TensorLRScheduler 

pub struct TensorLRScheduler { /* private fields */ }

Learning rate scheduler with multiple strategy support.

Implementations

impl TensorLRScheduler

pub fn new(config: LRSchedulerConfig) -> Self

Create a new scheduler from the given config. The initial learning rate is computed for step 0.

pub fn step(&mut self) -> f64

Advance by one step, compute the new learning rate, and return it.

pub fn get_lr(&self) -> f64

Return the current learning rate without advancing the step counter.

pub fn set_step(&mut self, step: usize)

Jump to a specific step and recompute the learning rate.

pub fn remaining_steps(&self) -> Option<usize>

For finite schedules (Cosine, OneCycle, WarmupLinear) return the number of remaining steps. Returns None for schedules that run indefinitely.

pub fn reset(&mut self)

Reset to step 0.

pub fn stats(&self) -> LRSchedulerStats

Return a snapshot of the scheduler state.

pub fn config(&self) -> &LRSchedulerConfig

Read-only access to the config.

Trait Implementations

impl Clone for TensorLRScheduler

fn clone(&self) -> TensorLRScheduler

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for TensorLRScheduler

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

Formats the value using the given formatter.