TabularPredictor.set_model_best¶

TabularPredictor.set_model_best(model: str, save_trainer: bool = False)[source]¶

Sets the model to be used by default when calling predictor.predict(data). By default, this is the model with the best validation score, but this is not always the case. If manually set, this can be overwritten internally if further training occurs, such as through fit_extra, refit_full, or distill.

Parameters:

model (str) – Name of model to set to best. If model does not exist or cannot infer, raises an AssertionError.
save_trainer (bool, default = False) – If True, self._trainer is saved with the new model_best value, such that it is reflected when predictor is loaded in future from disk.