ray.train.UserCallback.after_report

UserCallback.after_report(run_context: TrainRunContext, metrics: List[Dict[str, Any]], checkpoint: Checkpoint | None)

Called after all workers have reported a metric + checkpoint via ray.train.report.

Parameters:

• run_context – The TrainRunContext for the current training run.

• metrics – A list of metric dictionaries reported by workers, where metrics[i] is the metrics dict reported by worker i.

• checkpoint – A Checkpoint object that has been persisted to storage. This is None if no workers reported a checkpoint (e.g. ray.train.report(metrics, checkpoint=None)).