capability__pretrain#

capability__pretrain()[source]#

Capability: the forecaster can use pretraining for global learning.

String name: "capability:pretrain"
Public capability tag
Values: boolean, True / False
Example: True
Default: False

The capability:pretrain tag indicates whether a forecaster supports pretraining on global/panel data before being fit to specific time series.

If the tag is True, the forecaster implements the pretrain method and can be used in the following workflow:

forecaster.pretrain(y_panel) - learn from panel/global data
forecaster.fit(y_series) - set context or fine-tune on specific series
forecaster.predict(fh) - make predictions

The pretrain method sets the forecaster state to "pretrained", and subsequent calls to fit will preserve the pretrained weights (enabling fine-tuning) rather than resetting the estimator.

If the tag is False, the forecaster does not support pretraining, and calling pretrain will have no effect.