ITrainingEpochReporter<T>
Interfaces · AiDotNet.Interfaces
Implemented by models that run their own epoch loop internally and can report each epoch back to the caller.
Known implementations: TimeSeriesModelBase<T>
For Beginners
Some models do all their training inside one call, so the library can’t see the individual passes (“epochs”) happening. This interface is how such a model tells the library about each pass, which is what makes progress reporting and early stopping work.
How It Works
The AiModelBuilder facade drives an epoch loop itself for models it trains step-by-step,
and reports every epoch to the callbacks registered via ConfigureTrainingCallback. Models
that instead own their loop (the TimeSeries family, whose Train the facade calls once)
are invisible to that machinery, so their users get no progress reporting and no early-stop
veto. Implementing this interface lets the facade hand its per-epoch hook to the model so those
features behave identically across both kinds of model.
Properties
| Property | Summary |
|---|---|
Func<TrainingProgress<T>, bool> TrainingEpochCallback { get; set; } |
Gets or sets the per-epoch hook invoked once at the end of every training epoch. |