You are looking at documentation for an older release. Not what you want? See the current release documentation.
No, data points in the training data are automatically split into training set and test set. After that, the trained model can be used to generate forecasts for the future with data points that were not previously included in training.