Beta 0.4.1

[ourownstory]

TBD

---

This discussion was created from the release Beta 0.4.1.

[rbeldagarcia]

Hello! Thanks for the update, some features are very useful like to allow create a global model with a unique data frame.

I write this comment to report some observation of the new version. I have tried to migrate my models written in 0.3.2 to 0.4.1 and I have found some issues.
In particular, these models are Global model with events, and the problems appear in:
Could you confirm if the requirements.txt is updated? I would like to validate that my environment is correct before reporting this as a bug. In particular, I have tried with:

pandas=1.3.5
neuralprophet=0.4.1
numpy=1.21.6

I have tried to create a new environment with requirements.txt but it reports a error trying to install torch>=1.8.0

df_utils.py (line 1006)

    for event in config_events.keys():
        event_feature = pd.Series([0.0] * df.shape[0])
        event_feature.index = df.index  #added to fix the errors, without this it reports NaN when the events are added to the dataframe in global model. 
        				 #The problem is that if df.index != event_feature.index then it return NaN when try to do: df[event] = event_feature


        # events_df may be None in case ID from original df is not provided in events df
        if events_df is None:
            dates = None
        else:
            dates = events_df[events_df.event == event].ds
            event_feature.iloc[df.ds.isin(dates)] = 1.0 #Added 'iloc' to support events that are only present in some 'ID' groups
        df[event] = event_feature
    return df

I attach you the code and a screen shoot

Neuralprophet041.zip

[ourownstory]

converted to issue #879