Hi, I followed the exact same steps mentioned in the tutorials for Long horizon forecast - https://nixtla.github.io/neuralforecast/examples/longhorizon_with_nhits.html It was working fine earlier but now I'm getting an error in this step:

nf.models[0].results.get_best_result().config

AttributeError: 'AutoNHITS' object has no attribute 'results'

Has there been any changes on this front recently as this was working earlier but now its not? Thanks.

Hey @Shreya Mathur, Here is the declaration of results: https://github.com/Nixtla/neuralforecast/blob/main/neuralforecast/common/_base_auto.py#L226

Yes there is a bug, thanks for finding it! The fitted models are now in models_fitted attribute, not in models. Can you index with the new name please? It has to be nf.models_fitted[0]...

Yes it works when I use

nf.models_fitted[0].results.get_best_result().config

Thanks for the quick reply!

it doesn't work for me with 1.6.3 version

Hi @Guillaume GALIE. We changed it back to

models

instead of

models_fitted

Ok thank you. will give a try. Is it possible to get the config from a saved model (.ckpt? )