fix encoder support for conformal models (#2627)

README.md

@@ -166,7 +166,7 @@ series.plot()
 * **Multivariate Support:** `TimeSeries` can be multivariate - i.e., contain multiple time-varying
   dimensions/columns instead of a single scalar value. Many models can consume and produce multivariate series.
 
-* **Multiple series training (global models):** All machine learning based models (incl. all neural networks)
+* **Multiple Series Training (Global Models):** All machine learning based models (incl. all neural networks)
   support being trained on multiple (potentially multivariate) series. This can scale to large datasets too.
 
 * **Probabilistic Support:** `TimeSeries` objects can (optionally) represent stochastic
@@ -177,10 +177,10 @@ series.plot()
 * **Conformal Prediction Support:** Our conformal prediction models allow to generate probabilistic forecasts with
   calibrated quantile intervals for any pre-trained global forecasting model.
 
-* **Past and Future Covariates support:** Many models in Darts support past-observed and/or future-known
+* **Past and Future Covariates Support:** Many models in Darts support past-observed and/or future-known
   covariate (external data) time series as inputs for producing forecasts.
 
-* **Static Covariates support:** In addition to time-dependent data, `TimeSeries` can also contain
+* **Static Covariates Support:** In addition to time-dependent data, `TimeSeries` can also contain
   static data for each dimension, which can be exploited by some models.
 
 * **Hierarchical Reconciliation:** Darts offers transformers to perform reconciliation.
@@ -189,7 +189,7 @@ series.plot()
 * **Regression Models:** It is possible to plug-in any scikit-learn compatible model
   to obtain forecasts as functions of lagged values of the target series and covariates.
 
-* **Training with sample weights:** All global models support being trained with sample weights. They can be
+* **Training with Sample Weights:** All global models support being trained with sample weights. They can be
   applied to each observation, forecasted time step and target column.
 
 * **Forecast Start Shifting:** All global models support training and prediction on a shifted output window.
@@ -198,7 +198,7 @@ series.plot()
 
 * **Explainability:** Darts has the ability to *explain* some forecasting models using Shap values.
 
-* **Data processing:** Tools to easily apply (and revert) common transformations on
+* **Data Processing:** Tools to easily apply (and revert) common transformations on
   time series data (scaling, filling missing values, differencing, boxcox, ...)
 
 * **Metrics:** A variety of metrics for evaluating time series' goodness of fit;

darts/models/forecasting/conformal_models.py

@@ -308,40 +308,6 @@ def predict(
             If `series` is given and is a sequence of several time series, this function returns
             a sequence where each element contains the corresponding `n` points forecasts.
         """
-        if series is None:
-            # then there must be a single TS, and that was saved in super().fit as self.training_series
-            if self.model.training_series is None:
-                raise_log(
-                    ValueError(
-                        "Input `series` must be provided. This is the result either from fitting on multiple series, "
-                        "or from not having fit the model yet."
-                    ),
-                    logger,
-                )
-            series = self.model.training_series
-
-        called_with_single_series = get_series_seq_type(series) == SeriesType.SINGLE
-
-        # guarantee that all inputs are either list of TimeSeries or None
-        series = series2seq(series)
-        if past_covariates is None and self.model.past_covariate_series is not None:
-            past_covariates = [self.model.past_covariate_series] * len(series)
-        if future_covariates is None and self.model.future_covariate_series is not None:
-            future_covariates = [self.model.future_covariate_series] * len(series)
-        past_covariates = series2seq(past_covariates)
-        future_covariates = series2seq(future_covariates)
-
-        super().predict(
-            n,
-            series,
-            past_covariates,
-            future_covariates,
-            num_samples,
-            verbose,
-            predict_likelihood_parameters,
-            show_warnings,
-        )
-
         # call predict to verify that all series have required input times
         _ = self.model.predict(
             n=n,
@@ -355,6 +321,10 @@ def predict(
             **kwargs,
         )
 
+        series = series or self.model.training_series
+        called_with_single_series = get_series_seq_type(series) == SeriesType.SINGLE
+        series = series2seq(series)
+
         # generate only the required forecasts for calibration (including the last forecast which is the output of
         # `predict()`)
         cal_start, cal_start_format = _get_calibration_hfc_start(

darts/tests/models/forecasting/test_conformal_model.py

@@ -1344,7 +1344,7 @@ def test_too_short_input_predict(self, config):
         else:
             # if `past_covariates` are too short, then it raises error from the forecasting_model.predict()
             assert str(exc.value).startswith(
-                "The `past_covariates` at list/sequence index 0 are not long enough."
+                "The `past_covariates` are not long enough."
             )
 
     @pytest.mark.parametrize(
@@ -1658,3 +1658,42 @@ def test_calibration_hfc_start_value_hist_fc(self, config):
             start=start,
             start_format="value",
         ) == (start_expected, "value")
+
+    def test_encoders(self):
+        """Tests support of covariates encoders."""
+        n = OUT_LEN + 1
+        min_length = IN_LEN + n
+
+        # create non-overlapping train and val series
+        series = tg.linear_timeseries(length=min_length)
+        val_series = tg.linear_timeseries(
+            start=series.end_time() + series.freq, length=min_length
+        )
+
+        model = train_model(
+            series,
+            model_params={
+                "lags_future_covariates": (IN_LEN, OUT_LEN),
+                "add_encoders": {"datetime_attribute": {"future": ["hour"]}},
+            },
+        )
+
+        cp_model = ConformalNaiveModel(model, quantiles=q)
+        assert (
+            cp_model.model.encoders is not None
+            and cp_model.model.encoders.encoding_available
+        )
+        assert model.uses_future_covariates
+
+        # predict: encoders using stored train series must work
+        _ = cp_model.predict(n=n)
+        # predict: encoding of new series without train overlap must work
+        _ = cp_model.predict(n=n, series=val_series)
+
+        # check the same for hfc
+        _ = cp_model.historical_forecasts(
+            forecast_horizon=n, series=series, overlap_end=True
+        )
+        _ = cp_model.historical_forecasts(
+            forecast_horizon=n, series=val_series, overlap_end=True
+        )