fix lags_past_covariates dict type breaks lags_future_covariates when output_chunk_shift>0

CHANGELOG.md

@@ -11,6 +11,8 @@ but cannot always guarantee backwards compatibility. Changes that may **break co
 
 **Improved**
 
+- Fixed a bug when initiating a `RegressionModel` with `lags_past_covariates` as dict and `lags_future_covariates` as some other type (not dict) and `output_chunk_shift>0`, [#2652
+](https://github.com/unit8co/darts/issues/2652) by [Jules Authier](https://github.com/authierj).
 - New model: `StatsForecastAutoTBATS`. This model offers the [AutoTBATS](https://nixtlaverse.nixtla.io/statsforecast/src/core/models.html#autotbats) model from Nixtla's `statsforecasts` library. [#2611](https://github.com/unit8co/darts/pull/2611) by [He Weilin](https://github.com/cnhwl).
 - Added the `title` attribute to `TimeSeries.plot()`. This allows to set a title for the plot. [#2639](https://github.com/unit8co/darts/pull/2639) by [Jonathan Koch](https://github.com/jonathankoch99).
 - Added parameter `component_wise` to `show_anomalies()` to separately plot each component in multivariate series. [#2544](https://github.com/unit8co/darts/pull/2544) by [He Weilin](https://github.com/cnhwl).

darts/models/forecasting/regression_model.py

@@ -393,7 +393,9 @@ def _generate_lags(
                 processed_lags[lags_abbrev] = [
                     lag_ + output_chunk_shift for lag_ in processed_lags[lags_abbrev]
                 ]
-                if processed_component_lags:
+                if processed_component_lags and list(tmp_components_lags.keys()) != [
+                    "default_lags"
+                ]:
                     processed_component_lags[lags_abbrev] = {
                         comp_: [lag_ + output_chunk_shift for lag_ in lags_]
                         for comp_, lags_ in processed_component_lags[