Follow-up to https://github.com/Nixtla/neuralforecast/issues/1193. Models generally have a limited forecasting horizon, and using longer horizons requires autoregressive forecasts. Currently, this has to be implemented manually by the user. Automating this would be really convenient. This could be either a separate function, or extension to .predict() method.
Description
Follow-up to https://github.com/Nixtla/neuralforecast/issues/1193. Models generally have a limited forecasting horizon, and using longer horizons requires autoregressive forecasts. Currently, this has to be implemented manually by the user. Automating this would be really convenient. This could be either a separate function, or extension to
.predict()method.Use case
Any long-horizon forecast.