BacktestBatchForecasterMixin

class openstef_beam.backtesting.backtest_forecaster.mixins.BacktestBatchForecasterMixin

Bases: object

Extension mixin for forecasters that support batch prediction operations.

Enables efficient processing of multiple prediction requests simultaneously, which can significantly improve performance for models that benefit from batch operations (e.g., neural networks, GPU-accelerated models).

Variables:

batch_size – Maximum number of predictions to process in a single batch.

batch_size: int | None = FieldInfo(annotation=NoneType, required=True, description='Batch size for prediction.')

predict_batch(batch: list[RestrictedHorizonVersionedTimeSeries]) → Sequence[TimeSeriesDataset | None]

Handles batch prediction across multiple HorizonTransform instances.

Parameters:

• batch (list[RestrictedHorizonVersionedTimeSeries]) – List of HorizonTransform instances for batch prediction.

• batch

Returns:

List of DataFrames with predictions or None for each instance.

Return type:

Sequence[TimeSeriesDataset | None]

predict_batch_versioned(batch: list[RestrictedHorizonVersionedTimeSeries]) → Sequence[TimeSeriesDataset | None]

Predicts a batch of versioned time series with the model.

Parameters:

• batch (list[RestrictedHorizonVersionedTimeSeries]) – List of HorizonTransform instances for batch prediction.

• batch

Returns:

List of VersionedTimeSeriesDataset with predictions.

Return type:

Sequence[TimeSeriesDataset | None]