How to predict in autoregressive manner?

[mtomic123]

I trained a DeepVAR estimator with prediction length = 24 and it is unclear to me how does one predict in autoregressive manner.

I used the following code to make predictions for my multivariate ds_test hourly dataset:

forecast_it, ts_it = make_evaluation_predictions(
    dataset=ds_test,  # test dataset
    predictor=predictor,  # predictor
    num_samples=100,  # number of sample paths we want for evaluation
)

however this seems to only make predictions for the last 24 hours of ds_test. I basically want to
make predictions for all rows in ds_test. In fact, I am mostly interested in one-step ahead predictions but in autoregressive manner where I only predict the next point while using known previous points.