search

sktime / pytorch-forecasting

Time series forecasting with PyTorch
https://pytorch-forecasting.readthedocs.io/
MIT License
4k stars 632 forks source link

[DOC] N_BEATS "AssertionError: filters should not remove entries all entries" when forecasting with new data - Check clarification of docs #766

Open fornasari12 opened 2 years ago

fornasari12 commented 2 years ago

Expected behavior

After training N-BEATS model with NBeats.from_dataset() I want to predict with an out of sample dataframe with the length of max_encoder_lentgh.

Actual behavior

It throws "AssertionError: filters should not remove entries all entries - check encoder/decoder lengths and lags" if the length of the dataframe is lower than: max_encoder_length + max_prediction_length.

Code to reproduce the problem

https://gist.github.com/fornasari12/5ae6dc414f730beb2bd5c8e267b6c9f1

fornasari12 commented 2 years ago

hello @jdb78, could you help me with this error?

pnsafari commented 2 years ago

Hi @fornasari12 did you find any solution?

benHeid commented 1 month ago

You need to specify an input dataset with the length of max_encoder_length + prediction_length. However, the last prediction_length values are ignored and you can mask them with 0.

TODOs: