Struct TrainingComponents 

pub struct TrainingComponents<M>
where
    M: LearnerModel,
{
    pub label: Option<String>,
    pub num_epochs: usize,
    pub checkpoint: Option<usize>,
    pub checkpointer: Option<LearningCheckpointer<M>>,
    pub grad_accumulation: Option<usize>,
    pub interrupter: Interrupter,
    pub early_stopping: Option<Box<dyn CloneEarlyStoppingStrategy>>,
    pub event_processor: AsyncProcessorTraining<LearnerEvent<<M as TrainStep>::Output>, LearnerEvent<<M as InferenceStep>::Output>>,
    pub event_store: Arc<EventStoreClient>,
    pub summary: Option<LearnerSummaryConfig>,
}

Struct to minimise parameters passed to SupervisedLearningStrategy::train. These components are used during training.

Fields

label: Option<String>

An optional label for this training.

num_epochs: usize

The total number of epochs

checkpoint: Option<usize>

The epoch number from which to continue the training.

checkpointer: Option<LearningCheckpointer<M>>

A checkpointer used to load and save learner checkpoints.

grad_accumulation: Option<usize>

Enables gradients accumulation.

interrupter: Interrupter

An Interupter that allows aborting the training/evaluation process early.

early_stopping: Option<Box<dyn CloneEarlyStoppingStrategy>>

Cloneable reference to an early stopping strategy.

event_processor: AsyncProcessorTraining<LearnerEvent<<M as TrainStep>::Output>, LearnerEvent<<M as InferenceStep>::Output>>

An EventProcessor that processes events happening during training and validation.

event_store: Arc<EventStoreClient>

A reference to an EventStoreClient.

summary: Option<LearnerSummaryConfig>

Config for creating a summary of the learning