Feat/auto-regression and future values of past covariates documentation #2049
Conversation
Codecov ReportAll modified and coverable lines are covered by tests ✅
❗ Your organization needs to install the Codecov GitHub app to enable full functionality. Additional details and impacted files@@ Coverage Diff @@
## master #2049 +/- ##
==========================================
+ Coverage 93.82% 93.85% +0.03%
==========================================
Files 134 134
Lines 13091 13109 +18
==========================================
+ Hits 12282 12303 +21
+ Misses 809 806 -3 ☔ View full report in Codecov by Sentry. |
There was a problem hiding this comment.
Thanks for this @madtoinou 🚀
As discussed, it would be nice if the warning is raised once only for historical forecasting, but if it's too tedious, we can leave it like that.
…hen using historical forecast with a global model
There was a problem hiding this comment.
Thanks @madtoinou , looks good as well 🚀
Had some minor suggestions and we should probably merge the fit/predict_kwargs PR first.
| @@ -328,6 +328,7 @@ def _predict_wrapper( | |||
| num_samples: int, | |||
| verbose: bool = False, | |||
| predict_likelihood_parameters: bool = False, | |||
| show_warnings: bool = True, | |||
There was a problem hiding this comment.
we should probably merge the fit/predict kwargs for hist_fc before this PR and then adapt it here, right?
There was a problem hiding this comment.
It looks like regression models never call the super().predict() and this warning would never be raised when going through the optimized routine. Can we add this somehow?
Co-authored-by: Dennis Bader <[email protected]>
…d torch models, updated docstring about past/future covariates requirements in the torch models that were missing it
There was a problem hiding this comment.
Looks great, thanks @madtoinou 🚀
I updated the docs a bit for torch and regression models to have a more unified style for input/output chunk length and covariates support
Fixes #1822, fixes #2005.
Summary
GlobalForecastingModel.predict()(class where output_chunk is defined) whenn > output_chunk_lengthand the model uses past covariates.output_chunk_lengthfor theRegressionModelsto mention auto-regression and the usage of the past covariatesOther Information
The covariates section of the User guide already contain the following passage :
Should it be more emphasized? Modified?