TrajectoryMetricsReporter
python
TrajectoryMetricsReporter(store: TrajectoryStore | None, configuration: TrajectoryTrainingConfiguration | None, metrics: MetricsTracker) -> NoneEmits capture, exploration, repository, and utility metrics for one trajectory store.
Initialize the reporter.
Parameters:
store(TrajectoryStore | None) – The trajectory store to report on, orNonewhen trajectory training is disabled.configuration(TrajectoryTrainingConfiguration | None) – Optional trajectory training configuration.metrics(MetricsTracker) – Metrics tracker.
Methods
log_capture_metrics
python
log_capture_metrics(step: int, *, recorded: int, stored: int, dropped: int, replaced: int) -> NoneEmit capture and occupancy metrics.
The body scans the store for summaries, so it is skipped entirely while the metric time gate is closed; the writes would only be held as pending values anyway.
Parameters:
step(int) – The global step for metrics.recorded(int) – Total interesting segments recorded so far.stored(int) – Total segments stored so far.dropped(int) – Total segments dropped so far.replaced(int) – Total segments that replaced an existing entry so far.