Add supports_multivariate property to ForecastingModel

[felixdivo]

Fixes #1844.

Summary

The property was added to all models.

Other Information

I just assumed that this property then holds for all acpects of the time series, i.e. targets, covariates, ...

[felixdivo]

CC @madtoinou

[madtoinou]

Looking nice, thank you for also fixing the type hinting for some method.

In order to slightly simplify the code, would it be possible to define the property as returning True for the GlobalForecastingModels while keeping what you did for the LocalForecastingModels?

PS: it appears that you forgot to redefine the property for the FourTheta model in "theta.py"

[codecov-commenter]

Codecov Report

Patch coverage: 69.90% and project coverage change: -0.23 ⚠️

Comparison is base (6d373f4) 94.16% compared to head (efc0344) 93.94%.

❗ Your organization is not using the GitHub App Integration. As a result you may experience degraded service beginning May 15th. Please install the Github App Integration for your organization. Read more.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #1848      +/-   ##
==========================================
- Coverage   94.16%   93.94%   -0.23%     
==========================================
  Files         125      125              
  Lines       11642    11724      +82     
==========================================
+ Hits        10963    11014      +51     
- Misses        679      710      +31     

Impacted Files	Coverage Δ
darts/models/forecasting/arima.py	91.48% <66.66%> (-1.70%)	⬇️
darts/models/forecasting/auto_arima.py	96.66% <66.66%> (-3.34%)	⬇️
darts/models/forecasting/baselines.py	96.29% <66.66%> (-3.71%)	⬇️
darts/models/forecasting/block_rnn_model.py	96.49% <66.66%> (-1.70%)	⬇️
darts/models/forecasting/croston.py	90.24% <66.66%> (-1.87%)	⬇️
darts/models/forecasting/dlinear.py	99.11% <66.66%> (-0.89%)	⬇️
darts/models/forecasting/ensemble_model.py	95.83% <66.66%> (-0.75%)	⬇️
darts/models/forecasting/exponential_smoothing.py	97.91% <66.66%> (-2.09%)	⬇️
darts/models/forecasting/fft.py	95.57% <66.66%> (-0.79%)	⬇️
darts/models/forecasting/kalman_forecaster.py	97.29% <66.66%> (-2.71%)	⬇️
... and 18 more

... and 4 files with indirect coverage changes

☔ View full report in Codecov by Sentry.
📢 Do you have feedback about the report comment? Let us know in this issue.

[felixdivo]

@madtoinou this should be done now.

[madtoinou]

Looking good! Some minors comments before this is ready for merging.

[madtoinou]

Added suggestions for the docstrings of some RegressionModel, to make it more explicit that darts wrap around these models to make them multivariate when necessary.

On the contrary, RandomForest supports multivariate out of the box, same for XGBoost for example (so no need to mention anything in the docstring).

[dennisbader]

Looks really good @felixdivo , thanks a lot 🚀

Only had some minor comments:

• Linting checks fail because of the added docstring for RegressionModels.
• we can let RegressionModel handle the multivariate support and drop the properties in the sub classes

[felixdivo]

Done :)

[dennisbader]

Perfect, thanks a lot @felixdivo 🚀 💯