fix pass verbose to fit in historical_forecast

CHANGELOG.md

@@ -12,6 +12,9 @@ but cannot always guarantee backwards compatibility. Changes that may **break co
 **Improved**
 
 **Fixed**
+- Added parameter `verbose` to `ForecastingModel.fit()` that allows to control the verbosity for model fitting. Ignored if the underlying model does not support it. [#2805](https://github.com/unit8co/darts/pull/2805) by [Timon Erhart](https://github.com/turbotimon).
+- Fixed a bug in `historical_forecast` where the `verbose` parameter was not propagated to the the model's `fit()` method. [#2805](https://github.com/unit8co/darts/pull/2805) by [Timon Erhart](https://github.com/turbotimon).
+2805) by [Timon Erhart](https://github.com/turbotimon).
 
 **Dependencies**
 

darts/models/forecasting/baselines.py

@@ -51,7 +51,7 @@ def __init__(self):
     def supports_multivariate(self) -> bool:
         return True
 
-    def fit(self, series: TimeSeries):
+    def fit(self, series: TimeSeries, verbose: bool = False):
         super().fit(series)
 
         self.mean_val = np.mean(series.values(copy=False), axis=0)
@@ -112,7 +112,7 @@ def supports_multivariate(self) -> bool:
     def min_train_series_length(self):
         return max(self.K, 3)
 
-    def fit(self, series: TimeSeries):
+    def fit(self, series: TimeSeries, verbose: bool = False):
         super().fit(series)
 
         raise_if_not(
@@ -167,7 +167,7 @@ def __init__(self):
     def supports_multivariate(self) -> bool:
         return True
 
-    def fit(self, series: TimeSeries):
+    def fit(self, series: TimeSeries, verbose: bool = False):
         super().fit(series)
         assert series.n_samples == 1, "This model expects deterministic time series"
 
@@ -235,7 +235,7 @@ def min_train_series_length(self):
     def __str__(self):
         return f"NaiveMovingAverage({self.input_chunk_length})"
 
-    def fit(self, series: TimeSeries):
+    def fit(self, series: TimeSeries, verbose: bool = False):
         super().fit(series)
         raise_if_not(
             series.is_deterministic,
@@ -332,6 +332,7 @@ def fit(
         past_covariates: Optional[Union[TimeSeries, Sequence[TimeSeries]]] = None,
         future_covariates: Optional[Union[TimeSeries, Sequence[TimeSeries]]] = None,
         sample_weight: Optional[Union[TimeSeries, Sequence[TimeSeries], str]] = None,
+        verbose: bool = False,
     ):
         super().fit(
             series=series,
@@ -351,6 +352,7 @@ def fit(
                     sample_weight=sample_weight
                     if model.supports_sample_weight
                     else None,
+                    verbose=verbose,
                 )
         return self
 

darts/models/forecasting/ensemble_model.py

@@ -200,6 +200,7 @@ def fit(
         series: Union[TimeSeries, Sequence[TimeSeries]],
         past_covariates: Optional[Union[TimeSeries, Sequence[TimeSeries]]] = None,
         future_covariates: Optional[Union[TimeSeries, Sequence[TimeSeries]]] = None,
+        verbose: bool = False,
     ):
         """
         Fits the model on the provided series.

darts/models/forecasting/exponential_smoothing.py

@@ -104,7 +104,7 @@ def __init__(
         self.fit_kwargs = fit_kwargs
         self.model = None
 
-    def fit(self, series: TimeSeries):
+    def fit(self, series: TimeSeries, verbose: bool = False):
         super().fit(series)
         self._assert_univariate(series)
         series = self.training_series

darts/models/forecasting/fft.py

@@ -303,7 +303,7 @@ def _null_trend(self, x) -> Callable:
         """Helper function, used to make FFT model pickable."""
         return 0
 
-    def fit(self, series: TimeSeries):
+    def fit(self, series: TimeSeries, verbose: bool = False):
         series = fill_missing_values(series)
         super().fit(series)
         self._assert_univariate(series)

darts/models/forecasting/forecasting_model.py

@@ -169,13 +169,15 @@ def __init__(self, *args, **kwargs):
         self.encoders = self.initialize_encoders(default=True)
 
     @abstractmethod
-    def fit(self, series: TimeSeries) -> "ForecastingModel":
+    def fit(self, series: TimeSeries, verbose=False) -> "ForecastingModel":
         """Fit/train the model on the provided series.
 
         Parameters
         ----------
         series
             A target time series. The model will be trained to forecast this time series.
+        verbose
+            Optionally, set the fit verbosity. Not effective for all models.
 
         Returns
         -------
@@ -1188,6 +1190,7 @@ def retrain_func(
                             past_covariates=past_covariates_,
                             future_covariates=future_covariates_,
                             sample_weight=sample_weight_,
+                            verbose=verbose,
                             **fit_kwargs,
                         )
                     else:
@@ -2950,7 +2953,7 @@ def _model_encoder_settings(
         return None, None, False, False, None, None
 
     @abstractmethod
-    def fit(self, series: TimeSeries) -> "LocalForecastingModel":
+    def fit(self, series: TimeSeries, verbose: bool = False) -> "LocalForecastingModel":
         super().fit(series)
         series._assert_deterministic()
 
@@ -3011,6 +3014,7 @@ def fit(
         series: Union[TimeSeries, Sequence[TimeSeries]],
         past_covariates: Optional[Union[TimeSeries, Sequence[TimeSeries]]] = None,
         future_covariates: Optional[Union[TimeSeries, Sequence[TimeSeries]]] = None,
+        verbose: bool = False,
     ) -> "GlobalForecastingModel":
         """Fit/train the model on (potentially multiple) series.
 
@@ -3033,6 +3037,8 @@ def fit(
             be used by some models as an input. The covariate(s) may or may not be multivariate, but if multiple
             covariates are provided they must have the same number of components. If `future_covariates` is provided,
             it must contain the same number of series as `series`.
+        verbose
+            Optionally, set the fit verbosity. Not effective for all models.
 
         Returns
         -------
@@ -3237,7 +3243,12 @@ class FutureCovariatesLocalForecastingModel(LocalForecastingModel, ABC):
     All implementations must implement the :func:`_fit()` and :func:`_predict()` methods.
     """
 
-    def fit(self, series: TimeSeries, future_covariates: Optional[TimeSeries] = None):
+    def fit(
+        self,
+        series: TimeSeries,
+        future_covariates: Optional[TimeSeries] = None,
+        verbose: bool = False,
+    ):
         """Fit/train the model on the (single) provided series.
 
         Optionally, a future covariates series can be provided as well.
@@ -3250,6 +3261,8 @@ def fit(self, series: TimeSeries, future_covariates: Optional[TimeSeries] = None
             A time series of future-known covariates. This time series will not be forecasted, but can be used by
             some models as an input. It must contain at least the same time steps/indices as the target `series`.
             If it is longer than necessary, it will be automatically trimmed.
+        verbose
+            Optionally, set the fit verbosity. Not effective for all models.
 
         Returns
         -------

darts/models/forecasting/lgbm.py

@@ -259,6 +259,7 @@ def fit(
         val_sample_weight: Optional[
             Union[TimeSeries, Sequence[TimeSeries], str]
         ] = None,
+        verbose: bool = False,
         **kwargs,
     ):
         """
@@ -300,6 +301,8 @@ def fit(
             are extracted from the end of the global weights. This gives a common time weighting across all series.
         val_sample_weight
             Same as for `sample_weight` but for the evaluation dataset.
+        verbose
+            Optionally, set the fit verbosity. Not effective for all models.
          **kwargs
             Additional kwargs passed to `lightgbm.LGBRegressor.fit()`
         """
@@ -321,6 +324,7 @@ def fit(
                     n_jobs_multioutput_wrapper=n_jobs_multioutput_wrapper,
                     sample_weight=sample_weight,
                     val_sample_weight=val_sample_weight,
+                    verbose=verbose,
                     **kwargs,
                 )
                 # store the trained model in the container as it might have been wrapped by MultiOutputRegressor
@@ -338,6 +342,7 @@ def fit(
             n_jobs_multioutput_wrapper=n_jobs_multioutput_wrapper,
             sample_weight=sample_weight,
             val_sample_weight=val_sample_weight,
+            verbose=verbose,
             **kwargs,
         )
         return self

darts/models/forecasting/linear_regression_model.py

@@ -210,6 +210,7 @@ def fit(
         max_samples_per_ts: Optional[int] = None,
         n_jobs_multioutput_wrapper: Optional[int] = None,
         sample_weight: Optional[Union[TimeSeries, Sequence[TimeSeries], str]] = None,
+        verbose: bool = False,
         **kwargs,
     ):
         likelihood = self.likelihood
@@ -244,6 +245,7 @@ def fit(
                     max_samples_per_ts=max_samples_per_ts,
                     n_jobs_multioutput_wrapper=n_jobs_multioutput_wrapper,
                     sample_weight=sample_weight,
+                    verbose=verbose,
                     **kwargs,
                 )
                 # store the trained model in the container as it might have been wrapped by MultiOutputRegressor
@@ -262,6 +264,7 @@ def fit(
                 max_samples_per_ts=max_samples_per_ts,
                 n_jobs_multioutput_wrapper=n_jobs_multioutput_wrapper,
                 sample_weight=sample_weight,
+                verbose=verbose,
                 **kwargs,
             )
 

darts/models/forecasting/regression_ensemble_model.py

@@ -283,6 +283,7 @@ def fit(
         past_covariates: Optional[Union[TimeSeries, Sequence[TimeSeries]]] = None,
         future_covariates: Optional[Union[TimeSeries, Sequence[TimeSeries]]] = None,
         sample_weight: Optional[Union[TimeSeries, Sequence[TimeSeries], str]] = None,
+        verbose: bool = False,
     ):
         """
         Fits the forecasting models with the entire series except the last `regression_train_n_points` values, which
@@ -311,6 +312,8 @@ def fit(
             `"linear"` or `"exponential"` decay - the further in the past, the lower the weight. The weights are
             computed globally based on the length of the longest series in `series`. Then for each series, the weights
             are extracted from the end of the global weights. This gives a common time weighting across all series.
+        verbose
+            Optionally, set the fit verbosity. Not effective for all models.
         """
         super().fit(
             series, past_covariates=past_covariates, future_covariates=future_covariates
@@ -395,6 +398,7 @@ def fit(
                     sample_weight=sample_weight
                     if model.supports_sample_weight
                     else None,
+                    verbose=verbose,
                 )
 
         # we can call direct prediction in any case. Even if we overwrite with historical
@@ -442,6 +446,7 @@ def fit(
                     sample_weight=sample_weight
                     if model.supports_sample_weight
                     else None,
+                    verbose=verbose,
                 )
         return self
 

darts/models/forecasting/sklearn_model.py

@@ -55,6 +55,7 @@
   <https://unit8co.github.io/darts/examples/24-SKLearnClassifierModel-examples.html>`_
 """
 
+import inspect
 from abc import ABC, abstractmethod
 from collections import OrderedDict
 from collections.abc import Sequence
@@ -847,6 +848,7 @@ def _fit_model(
         val_past_covariates: Optional[Sequence[TimeSeries]] = None,
         val_future_covariates: Optional[Sequence[TimeSeries]] = None,
         val_sample_weight: Optional[Union[Sequence[TimeSeries], str]] = None,
+        verbose: bool = False,
         **kwargs,
     ):
         """
@@ -884,8 +886,16 @@ def _fit_model(
                     "`sample_weight` was ignored since underlying regression model's "
                     "`fit()` method does not support it."
                 )
+
+        # only pass "verbose" to the underlying model if it supports it
+        if "verbose" in inspect.signature(self.model.fit).parameters:
+            kwargs["verbose"] = verbose
+
         self.model.fit(
-            training_samples, training_labels, **sample_weight_kwargs, **kwargs
+            training_samples,
+            training_labels,
+            **sample_weight_kwargs,
+            **kwargs,
         )
 
         # generate and store the lagged components names (for feature importance analysis)
@@ -912,6 +922,7 @@ def fit(
         n_jobs_multioutput_wrapper: Optional[int] = None,
         sample_weight: Optional[Union[TimeSeries, Sequence[TimeSeries], str]] = None,
         stride: int = 1,
+        verbose: bool = False,
         **kwargs,
     ):
         """
@@ -949,6 +960,8 @@ def fit(
             The number of time steps between consecutive samples, applied starting from the end of the series. The same
             stride will be applied to both the training and evaluation set (if supplied and supported). This should be
             used with caution as it might introduce bias in the forecasts.
+        verbose
+            Optionally, set the fit verbosity. Not effective for all models.
         **kwargs
             Additional keyword arguments passed to the `fit` method of the model.
         """
@@ -1116,6 +1129,7 @@ def fit(
             val_sample_weight=val_sample_weight,
             max_samples_per_ts=max_samples_per_ts,
             stride=stride,
+            verbose=verbose,
             **kwargs,
         )
 
@@ -1864,6 +1878,7 @@ def _fit_model(
         future_covariates,
         max_samples_per_ts,
         sample_weight,
+        verbose: bool = False,
         **kwargs,
     ):
         """
@@ -1893,6 +1908,7 @@ def _fit_model(
             future_covariates=future_covariates,
             max_samples_per_ts=max_samples_per_ts,
             sample_weight=sample_weight,
+            verbose=verbose,
             **kwargs,
         )
 

darts/models/forecasting/theta.py

@@ -99,7 +99,7 @@ def __init__(
         if self.theta == 0:
             raise_log(ValueError("The parameter theta cannot be equal to 0."), logger)
 
-    def fit(self, series: TimeSeries):
+    def fit(self, series: TimeSeries, verbose: bool = False):
         super().fit(series)
         self._assert_univariate(series)
         ts = self.training_series
@@ -308,7 +308,7 @@ def __init__(
             logger,
         )
 
-    def fit(self, series):
+    def fit(self, series, verbose: bool = False):
         super().fit(series)
 
         self.length = len(series)

darts/models/forecasting/varima.py

@@ -125,7 +125,12 @@ def _differentiate_series(self, series: TimeSeries) -> TimeSeries:
             )
         return series
 
-    def fit(self, series: TimeSeries, future_covariates: Optional[TimeSeries] = None):
+    def fit(
+        self,
+        series: TimeSeries,
+        future_covariates: Optional[TimeSeries] = None,
+        verbose: bool = False,
+    ):
         # for VARIMA we need to process target `series` before calling
         # TransferableFutureCovariatesLocalForecastingModel's fit() method
         self._last_values = (

darts/models/forecasting/xgboost.py

@@ -262,6 +262,7 @@ def fit(
         val_sample_weight: Optional[
             Union[TimeSeries, Sequence[TimeSeries], str]
         ] = None,
+        verbose: bool = False,
         **kwargs,
     ):
         """
@@ -303,6 +304,8 @@ def fit(
             are extracted from the end of the global weights. This gives a common time weighting across all series.
         val_sample_weight
             Same as for `sample_weight` but for the evaluation dataset.
+        verbose
+            Optionally, set the fit verbosity. Not effective for all models.
         **kwargs
             Additional kwargs passed to `xgb.XGBRegressor.fit()`
         """
@@ -326,6 +329,7 @@ def fit(
                     n_jobs_multioutput_wrapper=n_jobs_multioutput_wrapper,
                     sample_weight=sample_weight,
                     val_sample_weight=val_sample_weight,
+                    verbose=verbose,
                     **kwargs,
                 )
                 # store the trained model in the container as it might have been wrapped by MultiOutputRegressor
@@ -343,6 +347,7 @@ def fit(
             n_jobs_multioutput_wrapper=n_jobs_multioutput_wrapper,
             sample_weight=sample_weight,
             val_sample_weight=val_sample_weight,
+            verbose=verbose,
             **kwargs,
         )
         return self