Fix/hfc opti reg prob (unit8co#2588)

* fix: check that model is probabilistic when num samples is greater than 1 for optimized historical forecasts

* feat: update the tests accordingly

* update changelog

* fix: simplify the test

* fix: remove typo

* fix: ignoring a linting commit for git blame

.git-blame-ignore-revs

@@ -6,3 +6,5 @@
 38cc6712a6f701703074a7a7c82ce0252fe869ee
 # Fix last isort issues and update Black to 22.1.0
 8158d3eaef9d9f6e04f219b029e306d1f1be46d5
+# Change Python target-version to 3.9 and update Ruff to 0.7.2
+18e2e3fd7d82d239ab24807fcc1033094ea09940

CHANGELOG.md

@@ -16,6 +16,7 @@ but cannot always guarantee backwards compatibility. Changes that may **break co
 **Fixed**
 
 - Fixed a bug when using `darts.utils.data.tabularization.create_lagged_component_names()` with target `lags=None`, that did not return any lagged target label component names. [#2576](https://github.com/unit8co/darts/pull/2576) by [Dennis Bader](https://github.com/dennisbader).
+- Fixed a bug when using `num_samples > 1` with a deterministic regression model and the optimized `historical_forecasts()` method, an exception was not raised. [#2576](https://github.com/unit8co/darts/pull/2588) by [Antoine Madrona](https://github.com/madtoinou).
 
 **Dependencies**
 

darts/tests/models/forecasting/test_regression_models.py

@@ -1289,6 +1289,48 @@ def test_historical_forecast(self, mode):
         )
         assert len(result) == 21
 
+    def test_opti_historical_forecast_predict_checks(self):
+        """
+        Verify that the sanity check implemented in ForecastingModel.predict are also defined for optimized historical
+        forecasts as it does not call this method
+        """
+        model = self.models[1](lags=5)
+
+        msg_expected = (
+            "The model has not been fitted yet, and `retrain` is ``False``. Either call `fit()` before "
+            "`historical_forecasts()`, or set `retrain` to something different than ``False``."
+        )
+        # untrained model, optimized
+        with pytest.raises(ValueError) as err:
+            model.historical_forecasts(
+                series=self.sine_univariate1,
+                start=0.9,
+                forecast_horizon=1,
+                retrain=False,
+                enable_optimization=True,
+                verbose=False,
+            )
+        assert str(err.value) == msg_expected
+
+        model.fit(
+            series=self.sine_univariate1,
+        )
+        # deterministic model, num_samples > 1, optimized
+        with pytest.raises(ValueError) as err:
+            model.historical_forecasts(
+                series=self.sine_univariate1,
+                start=0.9,
+                forecast_horizon=1,
+                retrain=False,
+                enable_optimization=True,
+                num_samples=10,
+                verbose=False,
+            )
+        assert (
+            str(err.value)
+            == "`num_samples > 1` is only supported for probabilistic models."
+        )
+
     @pytest.mark.parametrize(
         "config",
         [

darts/utils/historical_forecasts/utils.py

@@ -155,6 +155,13 @@ def _historical_forecasts_general_checks(model, series, kwargs):
                     logger,
                 )
 
+    # duplication of ForecastingModel.predict() check for the optimized historical forecasts implementations
+    if not model.supports_probabilistic_prediction and n.num_samples > 1:
+        raise_log(
+            ValueError("`num_samples > 1` is only supported for probabilistic models."),
+            logger,
+        )
+
     # check direct likelihood parameter prediction before fitting a model
     if n.predict_likelihood_parameters:
         if not model.supports_likelihood_parameter_prediction: