BatchEpisodeSampler<T, TInput, TOutput>

Models & Types · AiDotNet.MetaLearning.Data

Efficiently samples batches of episodes with optional prefetching and caching. Wraps an ITaskSampler and provides batch-level operations such as prefetching the next batch while the current batch is being trained on.

Constructors

Constructor Summary
BatchEpisodeSampler(ITaskSampler<T, TInput, TOutput> sampler, int batchSize, int prefetchCount) Creates a batch episode sampler.

Properties

Property Summary
int BatchSize { get; } Gets the number of episodes per batch.
int PrefetchedCount { get; } Gets the number of episodes currently in the prefetch buffer.

Methods

Method Summary
IReadOnlyList<IEpisode<T, TInput, TOutput>> NextBatch() Gets the next batch of episodes.
TaskBatch<T, TInput, TOutput> NextTaskBatch() Gets a task batch suitable for MetaTrain.
void ProvideFeedback(IReadOnlyList<IEpisode<T, TInput, TOutput>> episodes, IReadOnlyList<double> losses) Provides feedback for the sampled episodes (delegates to underlying sampler).