Timesfm2Model

timesfm2_model.Timesfm2Model · inherits BaseModel

TimesFM 2.5 foundation model wrapper with stochastic (quantile) output.

Uses the transformers TimesFm2_5ModelForPrediction which returns both mean_predictions and full_predictions (quantiles). Quantiles are converted to pseudo-samples for stochastic metrics (CRPS, etc.).

__init__(self, params, settings)

Initializes the TimesFM 2.5 model wrapper.

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

train(self, y_context, y_target, timestamps_context, timestamps_target, x_context=None, x_target=None, **kwargs) -> Timesfm2Model

Foundation model: no training needed. Build model and mark as fitted. Covariates (x_context, x_target) are ignored; TimesFM 2.5 has no native covariate support.

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: Timesfm2Model (the fitted model).

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

Make predictions using TimesFM 2.5. Returns stochastic samples (num_samples, forecast_horizon, num_targets) from quantile forecasts. Covariates are ignored.

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 (stochastic samples (num_samples, forecast_horizon, num_targets) from quantile forecasts).