Is that normal when I try to move DistributedMlForecast to local by calling method .to_local() I get an error -

mlforecast/distributed/forecast.py", line 795, in combine_target_tfms [part[i] for part in by_partition] for i in range(len(by_partition[0])) TypeError: object of type 'NoneType' has no len()

Which target transformations are you using?

RollingMean, SeasonalRollingMean. Same error I had with more complex using Combine and operator

Those are lag transforms, target transforms are like differences, scalers, etc

lag transforms

I see. The code assumes that if there aren't any target transforms they are an empty list but they're actually None, so that's what's failing. In the meantime, can you try adding a dummy global transformation? e.g.

from mlforecast.target_transforms import GlobalSklearnTransformer
from sklearn.preprocessing import FunctionTransformer

dummy_tfm = GlobalSklearnTransformer(FunctionTransformer())
fcst = DistributedMLForecast(..., target_transforms=[dummy_tfm])

That'll just apply the identity and should work while we fix the issue

Thanks, I will give it a try