TrainerBase<T>
Base Classes · AiDotNet.Training
Abstract base class for all trainers, providing shared infrastructure for configuration-driven training pipelines.
Implements: ITrainer<T>
Known derived types: Trainer<T>
For Beginners
Think of TrainerBase as a template for running experiments.
It handles all the boilerplate—loading data, setting up the model/optimizer/loss,
timing the run, and collecting results—so that concrete trainers only need to
implement the actual training strategy in TrainEpoch.
How It Works
The inheritance pattern follows the project’s architecture requirements:
ITrainer<T> (interface) → TrainerBase<T> (base) → Trainer<T> (concrete).
To create a custom training strategy, inherit from this base class and override
TrainEpoch.
Properties
| Property | Summary |
|---|---|
Action<string> LogAction { get; set; } |
Gets or sets the action used for logging training messages. |
TrainingRecipeConfig Config { get; } |
Methods
| Method | Summary |
|---|---|
Task<TrainingResult<T>> RunAsync() |
|
TrainingResult<T> Run() |
|
void SetData(Matrix<T> features, Vector<T> labels) |
Sets in-memory feature and label data for training, bypassing CSV loading. |