[Python] How can I implement single step prediction into N-Beats; pytorch forecasting

I am using N-Beats with pytorch forecasting for my own data. I split my training data into training (3599 Time Steps) and validation data (1566 Time Steps) and I must use single step prediction. I have 24 different Time Series. I am now confused how I can bring the single step prediction together with the validation horizont.

what i tried was to set the max_encoder_length to e.g. 150 and the max_prediction_length to 1 and the training_cutoff to 3599. What I got now as a result is one predicted time stamp for each TS. How can I get the other 1565 TS for the validation horizonts?! Thanks in advance!!

Continue reading...