Chronos2: Training data

After reading some articles it is still unclear to me why exactly the AutoGluon tutorial for Chronos2 creates training data. If I am not wrong, foundation models use transfer learning and the weights do not change anymore and fine tuning does not happen in this code, so why do I need to provide training data? AutoGluon describes it as:

As promised, Chronos does not take any time to fit. The fit call merely serves as a proxy for the TimeSeriesPredictor to do some of its chores under the hood, such as inferring the frequency of time series and saving the predictor’s state to disk.

Let’s use the predict method to generate forecasts.

https://auto.gluon.ai/dev/tutorials/timeseries/forecasting-chronos.html

num_test_windows = 3 prediction_length = 48 train_data, test_data = data.train_test_split(num_test_windows * prediction_length) predictor = TimeSeriesPredictor(prediction_length=prediction_length).fit( train_data, presets="chronos2", )