SundialModel

chronax.models.sundial_model.SundialModel · inherits BaseModel

Sundial diffusion-based foundation model for time series forecasting.

Sundial generates multiple sample trajectories via diffusion, making it a stochastic forecaster. It operates on univariate sequences, so multivariate targets are handled by iterating over each target independently with z-score normalization.

__init__(self, params: Dict[str, Any], settings: Dict[str, Any])

Initializes the Sundial model wrapper.

Parameter	Type	Default	Description
params	Dict[str, Any]	-	(undocumented)
settings	Dict[str, Any]	-	(undocumented)

train(self, y_context: np.ndarray, y_target: np.ndarray, timestamps_context: np.ndarray, timestamps_target: np.ndarray, x_context: Optional[np.ndarray] = None, x_target: Optional[np.ndarray] = None, **kwargs) -> SundialModel

Trains the Sundial model by loading the pre-trained Hugging Face model and setting it to evaluation mode.

Parameters:

Parameter	Type	Default	Description
y_context	np.ndarray	-	(undocumented)
y_target	np.ndarray	-	(undocumented)
timestamps_context	np.ndarray	-	(undocumented)
timestamps_target	np.ndarray	-	(undocumented)
x_context	Optional[np.ndarray]	None	(undocumented)
x_target	Optional[np.ndarray]	None	(undocumented)
**kwargs	-	-	(undocumented)

Returns: Self (the fitted forecaster).

predict(self, y_context: np.ndarray, timestamps_context: np.ndarray, timestamps_target: np.ndarray, x_context: Optional[np.ndarray] = None, x_target: Optional[np.ndarray] = None, **kwargs) -> np.ndarray

Generates sample trajectories for the forecast horizon using the fitted Sundial model.

Parameters:

Parameter	Type	Default	Description
y_context	np.ndarray	-	(undocumented)
timestamps_context	np.ndarray	-	(undocumented)
timestamps_target	np.ndarray	-	(undocumented)
x_context	Optional[np.ndarray]	None	(undocumented)
x_target	Optional[np.ndarray]	None	(undocumented)
**kwargs	-	-	(undocumented)

Returns: np.ndarray (An array of predicted samples, typically of shape (num_samples, forecast_horizon, num_targets)).

Raises:

Exception	Description
ValueError	If train() has not been called prior to prediction.