Struct ScheduledLSTMTrainer

pub struct ScheduledLSTMTrainer<L: LossFunction, O: Optimizer, S: LearningRateScheduler> {
    pub network: LSTMNetwork,
    pub loss_function: L,
    pub optimizer: ScheduledOptimizer<O, S>,
    pub config: TrainingConfig,
    pub metrics_history: Vec<TrainingMetrics>,
}

Specialized trainer for scheduled optimizers that automatically steps the scheduler

Fields

network: LSTMNetwork

loss_function: L

optimizer: ScheduledOptimizer<O, S>

config: TrainingConfig

metrics_history: Vec<TrainingMetrics>

Implementations

impl<L: LossFunction, O: Optimizer, S: LearningRateScheduler> ScheduledLSTMTrainer<L, O, S>

pub fn new( network: LSTMNetwork, loss_function: L, optimizer: ScheduledOptimizer<O, S>, ) -> Self

pub fn with_config(self, config: TrainingConfig) -> Self

pub fn train_sequence( &mut self, inputs: &[Array2<f64>], targets: &[Array2<f64>], ) -> f64

Train on a single sequence using backpropagation through time (BPTT)

pub fn train( &mut self, train_data: &[(Vec<Array2<f64>>, Vec<Array2<f64>>)], validation_data: Option<&[(Vec<Array2<f64>>, Vec<Array2<f64>>)]>, )

Train for multiple epochs with automatic scheduler stepping

pub fn evaluate(&mut self, data: &[(Vec<Array2<f64>>, Vec<Array2<f64>>)]) -> f64

Evaluate model performance on validation data

pub fn predict(&mut self, inputs: &[Array2<f64>]) -> Vec<Array2<f64>>

Generate predictions for input sequences

pub fn get_latest_metrics(&self) -> Option<&TrainingMetrics>

pub fn get_metrics_history(&self) -> &[TrainingMetrics]

pub fn set_training_mode(&mut self, training: bool)

Set network to training mode

pub fn get_current_lr(&self) -> f64

Get the current learning rate

pub fn get_current_epoch(&self) -> usize

Get the current epoch from the scheduler

pub fn reset_optimizer(&mut self)

Reset the optimizer and scheduler