Forecaster
Chronos2Forecaster
Interface to the Chronos-2 Zero-Shot Forecaster by Amazon Research.
Chronos-2 is a pretrained encoder-only time series foundation model developed by Amazon for zero-shot forecasting. It supports univariate, multivariate, and covariate-informed forecasting tasks within a single architecture. The official code and technical report are given at [1] and [2].
Unlike Chronos (v1), Chronos-2 natively handles multivariate targets, past-only covariates, and known-future covariates via a group attention mechanism described in [2].
Schnellstart
python
from sktime.forecasting.chronos2 import Chronos2Forecaster
estimator = Chronos2Forecaster(model_path: str='amazon/chronos-2', config: dict=None, seed: int | None=None, ignore_deps: bool=False)Parameter(4)
- model_pathstr, default=”amazon/chronos-2”
- Path to the Chronos-2 HuggingFace model.
- configdict, optional, default=None
Configuration overrides. Supported keys:
- “limit_prediction_length”bool, default=False
If True, raises an error when prediction_length exceeds the model’s maximum prediction length.
- seedint or None, optional, default=None
- Random seed for reproducibility.
- ignore_depsbool, optional, default=False
- If True, dependency checks are skipped.
Beispiele
>>> from sktime.datasets import load_airline
>>> from sktime.forecasting.chronos2 import Chronos2Forecaster
>>> from sktime.split import temporal_train_test_split
>>> y = load_airline ()
>>> y_train, y_test = temporal_train_test_split (y)
>>> forecaster = Chronos2Forecaster ("amazon/chronos-2")
>>> forecaster. fit (y_train)
>>> y_pred = forecaster. predict (fh = [1, 2, 3 ])