KronosForecaster

class KronosForecaster(model_path='NeoQuasar/Kronos-small', tokenizer_path='NeoQuasar/Kronos-Tokenizer-base', device='cpu', columns=None, freq='5min', start='2000-01-01', clip=5.0, predict_kwargs=None, deterministic=False)[source]

Kronos zero-shot forecaster for financial K-line/OHLC data.

This forecaster wraps Kronos [1], a foundation model for financial market data [2], through the sktime forecasting interface. This implementation is inference-only and uses the upstream KronosPredictor preprocessing and autoregressive inference path internally.

Parameters:

model_pathstr, default=”NeoQuasar/Kronos-small”: Hugging Face repository identifier or local path for the Kronos model. The default is the Kronos-small checkpoint [4]. Other released checkpoints include Kronos-mini [3] and Kronos-base [5].
tokenizer_pathstr, default=”NeoQuasar/Kronos-Tokenizer-base”: Hugging Face repository identifier or local path for the Kronos tokenizer. The default is the Kronos-Tokenizer-base checkpoint [6]. The released 2k tokenizer is also available [7].
devicestr, default=”cpu”: Device used for model and tokenizer inference.
columnslist of str or None, default=None: Optional positional mapping from columns in y to Kronos internal columns. Positions map to "open", "high", "low", "close", "volume", and "amount". If provided, at least the first four OHLC columns are required; volume and amount are optional. If None, literal open/high/low/close names are used when present, otherwise the first four numeric columns are used as OHLC. Literal volume/amount columns are used when present.
freqstr or pandas offset, default=”5min”: Frequency used to synthesize timestamps for Kronos when the training index is not a pd.PeriodIndex or pd.DatetimeIndex. The default is five minutes because Kronos is designed for financial K-line/OHLC data, where five-minute intraday bars are a common default granularity.
startstr or pd.Timestamp, default=”2000-01-01”: Start timestamp used with freq to synthesize timestamps for Kronos when the training index is not datetime-like. The default date is arbitrary; it provides deterministic calendar fields for Kronos while keeping the synthetic timestamp path independent of the original index.
clipfloat, default=5.0: Input normalization clipping value passed to KronosPredictor.
predict_kwargsdict or None, default=None: Additional keyword arguments passed directly to KronosPredictor.predict. Examples are T, top_k, top_p, sample_count, and verbose.
deterministicbool, default=False: Whether predictions should reset the PyTorch random seed before autoregressive sampling.

Attributes:

cutoff: Cut-off = “present time” state of forecaster.
fh: Forecasting horizon that was passed.
is_fitted: Whether fit has been called.
state: State of the estimator.

Notes

KronosForecaster expects financial K-line style data. The input y should contain OHLC columns, and may optionally contain volume and amount. For relative forecasting horizons with datetime-like indices, provide a sensible regular index/frequency so future timestamps line up with the data. If the calendar is irregular, pass an absolute forecasting horizon.

References

[1]

Kronos GitHub repository: https://github.com/shiyu-coder/Kronos

[2]

Lin, Z., Xia, Y., Liu, Z., Zhang, S., Wang, J., Yang, C., Dong, Q., Liu, H., Jiang, H., Wang, S., Xiong, X., and Zhao, B. (2025). Kronos: A Foundation Model for the Language of Financial Markets. arXiv. https://arxiv.org/abs/2508.02739

[3]

Kronos-mini model card: https://huggingface.co/NeoQuasar/Kronos-mini

[4]

Kronos-small model card: https://huggingface.co/NeoQuasar/Kronos-small

[5]

Kronos-base model card: https://huggingface.co/NeoQuasar/Kronos-base

[6]

Kronos-Tokenizer-base model card: https://huggingface.co/NeoQuasar/Kronos-Tokenizer-base

[7]

Kronos-Tokenizer-2k model card: https://huggingface.co/NeoQuasar/Kronos-Tokenizer-2k

Examples

>>> import pandas as pd
>>> from sktime.forecasting.base import ForecastingHorizon
>>> from sktime.forecasting.kronos import KronosForecaster
>>> url = (
...     "https://raw.githubusercontent.com/shiyu-coder/Kronos/"
...     "refs/heads/master/tests/data/regression_input.csv"
... )
>>> df = pd.read_csv(
...     url,
...     parse_dates=["timestamps"],
...     index_col="timestamps",
... )
>>> lookback, pred_len = 400, 120
>>> y = df.iloc[:lookback]
>>> fh = ForecastingHorizon(
...     df.index[lookback : lookback + pred_len],
...     is_relative=False,
... )
>>> forecaster = KronosForecaster(
...     model_path="NeoQuasar/Kronos-small",
...     tokenizer_path="NeoQuasar/Kronos-Tokenizer-base",
...     device="cpu",
...     deterministic=True,
...     predict_kwargs={
...         "T": 1.0,
...         "top_p": 0.9,
...         "sample_count": 1,
...         "verbose": True,
...     },
... )
>>> y_pred = forecaster.fit(y).predict(fh=fh)

Methods

`check_is_fitted`([method_name])	Check if the estimator has been fitted.
`clone`()	Obtain a clone of the object with same hyper-parameters and config.
`clone_tags`(estimator[, tag_names])	Clone tags from another object as dynamic override.
`create_test_instance`([parameter_set])	Construct an instance of the class, using first test parameter set.
`create_test_instances_and_names`([parameter_set])	Create list of all test instances and a list of names for them.
`fit`(y[, X, fh])	Fit forecaster to training data.
`fit_predict`(y[, X, fh, X_pred])	Fit and forecast time series at future horizon.
`get_class_tag`(tag_name[, tag_value_default])	Get class tag value from class, with tag level inheritance from parents.
`get_class_tags`()	Get class tags from class, with tag level inheritance from parent classes.
`get_config`()	Get config flags for self.
`get_fitted_params`([deep])	Get fitted parameters.
`get_param_defaults`()	Get object's parameter defaults.
`get_param_names`([sort])	Get object's parameter names.
`get_params`([deep])	Get a dict of parameters values for this object.
`get_pretrained_params`([deep])	Get pretrained parameters of this estimator.
`get_tag`(tag_name[, tag_value_default, ...])	Get tag value from instance, with tag level inheritance and overrides.
`get_tags`()	Get tags from instance, with tag level inheritance and overrides.
`get_test_params`([parameter_set])	Return testing parameter settings for the estimator.
`is_composite`()	Check if the object is composed of other BaseObjects.
`load_from_path`(serial)	Load object from file location.
`load_from_serial`(serial)	Load object from serialized memory container.
`predict`([fh, X])	Forecast time series at future horizon.
`predict_interval`([fh, X, coverage])	Compute/return prediction interval forecasts.
`predict_proba`([fh, X, marginal])	Compute/return fully probabilistic forecasts.
`predict_quantiles`([fh, X, alpha])	Compute/return quantile forecasts.
`predict_residuals`([y, X])	Return residuals of time series forecasts.
`predict_var`([fh, X, cov])	Compute/return variance forecasts.
`pretrain`(y[, X, fh])	Pre-train forecaster on panel (global) data.
`reset`()	Reset the object to a clean post-init state.
`save`([path, serialization_format])	Save serialized self to bytes-like object or to (.zip) file.
`score`(y[, X, fh])	Scores forecast against ground truth, using MAPE (non-symmetric).
`set_config`(**config_dict)	Set config flags to given values.
`set_params`(**params)	Set the parameters of this object.
`set_random_state`([random_state, deep, ...])	Set random_state pseudo-random seed parameters for self.
`set_tags`(**tag_dict)	Set instance level tag overrides to given values.
`update`(y[, X, update_params])	Update cutoff value and, optionally, fitted parameters.
`update_predict`(y[, cv, X, update_params, ...])	Make predictions and update model iteratively over the test set.
`update_predict_single`([y, fh, X, update_params])	Update model with new data and make forecasts.