Enum RLEvent 

pub enum RLEvent<TS, ES> {
    Start {
        total_items: usize,
        label: Option<String>,
    },
    TrainStep(EvaluationItem<TS>),
    EnvStep(EvaluationItem<ES>),
    EpisodeEnd(EvaluationItem<EpisodeSummary>),
    End(Option<LearnerSummary>),
}

Event happening during reinforcement learning.

Variants

Start

Signal the start of the process (e.g., learning starts).

Fields

total_items: usize

The total number of items to process during training (e.g., total number of environment steps).

label: Option<String>

An optional label for this training.

TrainStep(EvaluationItem<TS>)

Signal an agent’s training step.

EnvStep(EvaluationItem<ES>)

Signal a timestep of the agent-environment interface.

EpisodeEnd(EvaluationItem<EpisodeSummary>)

Signal an episode end.

End(Option<LearnerSummary>)

Signal the end of the process (e.g., learning ends).

Trait Implementations

impl<TS, ES> EventProcessorTraining<RLEvent<TS, ES>, AgentEvaluationEvent<ES>> for RLEventProcessor<TS, ES>

where TS: ItemLazy, ES: ItemLazy,

fn process_train( &mut self, event: RLEvent<TS, ES>, ) -> Result<(), EventProcessorError>

Collect a training event.

fn process_valid( &mut self, event: AgentEvaluationEvent<ES>, ) -> Result<(), EventProcessorError>

Collect a validation event.

fn renderer(self) -> Box<dyn MetricsRenderer>

Returns the renderer used for training.

fn flush(&mut self) -> Result<(), EventProcessorError>

Wait until previously submitted events are processed (no-op for sync processors).