Fix: Using gridsearch with use_fitted_values=True raises unexpected error

CHANGELOG.md

@@ -23,6 +23,7 @@ but cannot always guarantee backwards compatibility. Changes that may **break co
 - Fixed a bug in probabilistic `LinearRegressionModel.fit()`, where the `model` attribute was not pointing to all underlying estimators. [#2205](https://github.com/unit8co/darts/pull/2205) by [Antoine Madrona](https://github.com/madtoinou).
 - Raise an error in `RegressionEsembleModel` when the `regression_model` was created with `multi_models=False` (not supported). [#2205](https://github.com/unit8co/darts/pull/2205) by [Antoine Madrona](https://github.com/madtoinou).
 - Fixed a bug in `coefficient_of_variaton()` with `intersect=True`, where the coefficient was not computed on the intersection. [#2202](https://github.com/unit8co/darts/pull/2202) by [Antoine Madrona](https://github.com/madtoinou).
+- Fixed a bug in `gridsearch()` with `use_fitted_values=True`, where the model was not propely instantiated for sanity checks. [#2222](https://github.com/unit8co/darts/pull/2222) by [Antoine Madrona](https://github.com/madtoinou).
 
 ### For developers of the library:
 

darts/models/forecasting/forecasting_model.py

@@ -1483,10 +1483,28 @@ def gridsearch(
             logger,
         )
 
+        if not isinstance(parameters, dict):
+            raise_log(
+                ValueError(
+                    f"`parameters` should be a dictionary, received a: {type(parameters)}."
+                )
+            )
+
+        if not all(isinstance(params, list) for params in parameters.values()):
+            raise_log(
+                ValueError(
+                    "Every value in the `parameters` dictionary should be a list (possibly with only one element)."
+                ),
+                logger,
+            )
+
         if use_fitted_values:
             raise_if_not(
-                hasattr(model_class(), "fitted_values"),
-                "The model must have a fitted_values attribute to compare with the train TimeSeries",
+                hasattr(
+                    model_class(**{k: v[0] for k, v in parameters.items()}),
+                    "fitted_values",
+                ),
+                "The model must have a fitted_values attribute to compare with the train TimeSeries (local models)",
                 logger,
             )
 

darts/tests/models/forecasting/test_4theta.py

@@ -61,7 +61,7 @@ def test_best_model(self):
         linear_series = lt(length=50)
         series = sine_series + linear_series
         train_series, val_series = series.split_before(series.time_index[-10])
-        thetas = np.linspace(-3, 3, 30)
+        thetas = list(np.linspace(-3, 3, 30))
         best_model, _, _ = FourTheta.select_best_model(train_series, thetas)
         model = FourTheta(
             random.choice(thetas),