[BUG] Wrong import from statsforecast when installing u8darts-torch

[JeroenPeterBos]

Describe the bug
There are import errors when I try to import TFTModel after installing using u8darts-torch using conda.

To Reproduce
I run the following commands from an M1 macos device.

conda create -n bug-example -c conda-forge pytorch u8darts-torch -y
conda activate bug-example
python
>>> from darts.models import TFTModel

Which returns:

from statsforecast.models import croston_classic, croston_optimized, croston_sba
ImportError: cannot import name 'croston_classic' from 'statsforecast.models'

I checked statsforecast and indeed croston_classic does not exist, only _croston_classic and CrostonClassic.

Expected behavior
A clear and concise description of what you expected to happen.

System (please complete the following information):

• Python version: 3.8
• darts version: 0.20.0

Additional context
I am running this from an M1 mac with OS 12.5.1
Additionally, I first tried to install u8darts-all using conda create -n test python=3.8 pytorch u8darts-all, but that could not find any satisfable dependency configuration.

[JeroenPeterBos]

After some further digging I found that this issue is fixed in this PR that has already been merged:

If anyone else runs into this issue, you can probably solve it by explicitly installing statsforecasts==0.6.0.

[hrzn]

Yes, this should be fixed in 0.21.0 with statsforecasts>0.6.0 as well.