openstef 3.4.77__tar.gz → 3.4.79__tar.gz

{openstef-3.4.77 → openstef-3.4.79}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: openstef
-Version: 3.4.77
+Version: 3.4.79
 Summary: Open short term energy forecaster
 Home-page: https://github.com/OpenSTEF/openstef
 Author: Alliander N.V

{openstef-3.4.77 → openstef-3.4.79}/openstef/data_classes/prediction_job.py

@@ -140,6 +140,10 @@ class PredictionJobDataClass(BaseModel):
     data_prep_class: Optional[DataPrepDataClass] = Field(
         None, description="The import string for the custom data prep class"
     )
+    model_run_id: Optional[str] = Field(
+        None,
+        description="The specific model run number that should be used for the forecast. If not set, the latest model run will be used.",
+    )
 
     fallback_strategy: Optional[FallbackStrategy] = Field(
         FallbackStrategy.EXTREME_DAY,

{openstef-3.4.77 → openstef-3.4.79}/openstef/feature_engineering/holiday_features.py

@@ -1,8 +1,8 @@
 # SPDX-FileCopyrightText: 2017-2023 Contributors to the OpenSTEF project <korte.termijn.prognoses@alliander.com> # noqa E501>
 #
 # SPDX-License-Identifier: MPL-2.0
-"""This module contains all holiday related features."""
 from datetime import datetime, timedelta
+import collections
 
 import holidays
 import numpy as np
@@ -26,7 +26,6 @@ def generate_holiday_feature_functions(
         2022-12-24 - 2023-01-08 is the 'Kerstvakantie'
         2022-10-15 - 2022-10-23 is the 'HerfstvakantieNoord'
 
-
     The holidays are based on a manually generated csv file.
     The information is collected using:
     https://www.schoolvakanties-nederland.nl/ and the python holiday function
@@ -44,7 +43,6 @@ def generate_holiday_feature_functions(
         - Pinksteren
         - Kerst
 
-
     The 'Brugdagen' are updated untill dec 2020. (Generated using agenda)
 
     Args:
@@ -83,23 +81,34 @@ def generate_holiday_feature_functions(
             )
         }
     )
+
     # Define empty list to keep track of bridgedays
     bridge_days = []
-    # Loop over list of holidays names
+
+    # Group holiday dates by name
+    holiday_dates_by_name = collections.defaultdict(list)
     for date, holiday_name in sorted(country_holidays.items()):
-        # Define function explicitely to mitigate 'late binding' problem
-        def make_holiday_func(requested_date):
-            return lambda x: np.isin(x.index.date, np.array([requested_date]))
+        holiday_dates_by_name[holiday_name].append(date)
 
-        # Create lag function for each holiday
+    # Create one function per holiday name that checks all dates for that holiday
+    for holiday_name, dates in holiday_dates_by_name.items():
+        # Use a default argument to capture the dates at definition time
         holiday_functions.update(
-            {"is_" + holiday_name.replace(" ", "_").lower(): make_holiday_func(date)}
+            {
+                "is_"
+                + holiday_name.replace(
+                    " ", "_"
+                ).lower(): lambda x, dates_local=dates: np.isin(
+                    x.index.date, np.array(dates_local)
+                )
+            }
         )
 
-        # Check for bridge day
-        holiday_functions, bridge_days = check_for_bridge_day(
-            date, holiday_name, country_code, years, holiday_functions, bridge_days
-        )
+        # Check for bridge days for each date of this holiday
+        for date in dates:
+            holiday_functions, bridge_days = check_for_bridge_day(
+                date, holiday_name, country_code, years, holiday_functions, bridge_days
+            )
 
     # Add feature function that includes all bridgedays
     holiday_functions.update(
@@ -108,7 +117,7 @@ def generate_holiday_feature_functions(
 
     # Add school holidays if country is NL
     if country_code == "NL":
-        # Manully generated csv including all dutch schoolholidays for different regions
+        # Manually generated csv including all dutch schoolholidays for different regions
         df_holidays = pd.read_csv(path_to_school_holidays_csv, index_col=None)
         df_holidays["datum"] = pd.to_datetime(df_holidays.datum).apply(
             lambda x: x.date()
@@ -125,19 +134,17 @@ def generate_holiday_feature_functions(
 
         # Loop over list of holidays names
         for holiday_name in list(set(df_holidays.name)):
-            # Define function explicitely to mitigate 'late binding' problem
-            def make_holiday_func(holidayname=holiday_name):
-                return lambda x: np.isin(
-                    x.index.date,
-                    df_holidays.datum[df_holidays.name == holidayname].values,
-                )
-
-            # Create lag function for each holiday
+            # Use the holidayname as a default argument to capture it at definition time
             holiday_functions.update(
                 {
                     "is_"
-                    + holiday_name.replace(" ", "_").lower(): make_holiday_func(
-                        holidayname=holiday_name
+                    + holiday_name.replace(
+                        " ", "_"
+                    ).lower(): lambda x, holiday_name_local=holiday_name: np.isin(
+                        x.index.date,
+                        df_holidays.datum[
+                            df_holidays.name == holiday_name_local
+                        ].values,
                     )
                 }
             )
@@ -178,9 +185,10 @@ def check_for_bridge_day(
     if date in country_holidays:
         return holiday_functions, bridge_days
 
-    # Define function explicitely to mitigate 'late binding' problem
+    # Define function explicitly to mitigate 'late binding' problem
+    # Use a default argument to capture the date at definition time
     def make_holiday_func(requested_date):
-        return lambda x: np.isin(x.index.date, np.array([requested_date]))
+        return lambda x, dt=requested_date: np.isin(x.index.date, np.array([dt]))
 
     # Looking forward: If day after tomorow is a national holiday or
     # a saturday check if tomorow is not a national holiday

{openstef-3.4.77 → openstef-3.4.79}/openstef/model/regressors/median.py

@@ -304,9 +304,11 @@ class MedianRegressor(OpenstfRegressor, RegressorMixin):
 
         Which lag features are used is determined by the feature engineering step.
         """
-        feature_names, frequency, feature_to_lags_in_min = (
-            self._extract_and_validate_lags(x)
-        )
+        (
+            feature_names,
+            frequency,
+            feature_to_lags_in_min,
+        ) = self._extract_and_validate_lags(x)
 
         self.feature_names_ = list(feature_names)
         self.frequency_ = frequency

{openstef-3.4.77 → openstef-3.4.79}/openstef/model/serializer.py

@@ -18,6 +18,7 @@ from mlflow.store.artifact.artifact_repository_registry import get_artifact_repo
 from xgboost import XGBModel  # Temporary for backward compatibility
 
 from openstef.data_classes.model_specifications import ModelSpecificationDataClass
+from openstef.data_classes.prediction_job import PredictionJobDataClass
 from openstef.logging.logger_factory import get_logger
 from openstef.metrics.reporter import Report
 from openstef.model.regressors.regressor import OpenstfRegressor
@@ -143,20 +144,30 @@ class MLflowSerializer:
     def load_model(
         self,
         experiment_name: str,
+        model_run_id: Optional[str] = None,
     ) -> tuple[OpenstfRegressor, ModelSpecificationDataClass]:
-        """Load sklearn compatible model from MLFlow.
+        """ Load an sklearn-compatible model from MLflow.
 
+        This method retrieves a trained model and its specifications from MLflow 
+        based on the provided PredictionJobDataClass instance. It supports loading 
+        a specific model run if a run number is provided.
+            
         Args:
-            experiment_name: Name of the experiment, often the id of the predition job.
+                experiment_name (str): Name of the experiment, often the id of the predition job.
+                model_run_id (Optional[str]): The specific model run number that should be used for the forecast.
+
+        Returns:
+            tuple[OpenstfRegressor, ModelSpecificationDataClass]: A tuple containing 
+                the loaded model and its specifications.
 
-        Raises:
-            LookupError: If model is not found in MLflow.
+            LookupError: If the model is not found in MLflow or if an error occurs 
+                during the loading process.
 
         """
         try:
             models_df = self._find_models(
-                self.experiment_name_prefix + experiment_name, max_results=1
-            )  # return the latest finished run of the model
+                self.experiment_name_prefix + experiment_name, max_results=1, model_run_id=model_run_id)
+             # return the latest finished run of the model
             if not models_df.empty:
                 latest_run = models_df.iloc[0]  # Use .iloc[0] to only get latest run
             else:
@@ -172,7 +183,7 @@ class MLflowSerializer:
             )  # Path without file:///
             self.logger.info("Model successfully loaded with MLflow")
             return loaded_model, model_specs
-        except (AttributeError, MlflowException, OSError) as exception:
+        except (AttributeError, MlflowException, OSError) as exception:            
             raise LookupError("Model not found. First train a model!") from exception
 
     def get_model_age(
@@ -205,8 +216,13 @@ class MLflowSerializer:
         experiment_name: str,
         max_results: Optional[int] = 100,
         filter_string: str = "attribute.status = 'FINISHED'",
+        model_run_id: Optional[int] = None,
     ) -> pd.DataFrame:
         """Finds trained models for specific experiment_name sorted by age in descending order."""
+
+        if model_run_id is not None:
+            filter_string += f" AND attributes.run_id = '{model_run_id}'"
+
         models_df = mlflow.search_runs(
             experiment_names=[experiment_name],
             max_results=max_results,

{openstef-3.4.77 → openstef-3.4.79}/openstef/pipeline/create_forecast.py

@@ -52,11 +52,10 @@ def create_forecast_pipeline(
     # Use the alternative forecast model if it's specify in the pj
     if pj.alternative_forecast_model_pid:
         prediction_model_pid = pj.alternative_forecast_model_pid
-
     # Load most recent model for the given pid
     model, model_specs = MLflowSerializer(
         mlflow_tracking_uri=mlflow_tracking_uri
-    ).load_model(experiment_name=str(prediction_model_pid))
+    ).load_model(experiment_name=str(prediction_model_pid), model_run_id=pj.get("model_run_id"))
     return create_forecast_pipeline_core(pj, input_data, model, model_specs)
 
 

{openstef-3.4.77 → openstef-3.4.79}/openstef/pipeline/train_model.py

@@ -52,6 +52,7 @@ def train_model_pipeline(
         check_old_model_age: Check if training should be skipped because the model is too young
         mlflow_tracking_uri: Tracking URI for MLFlow
         artifact_folder: Path where artifacts, such as trained models, are stored
+        ignore_existing_models: If True, a new model is trained as if no old model exists.
 
     Returns:
         If pj.save_train_forecasts is False, None is returned
@@ -168,6 +169,7 @@ def train_model_pipeline_core(
         input_data: Input data
         old_model: Old model to compare to. Defaults to None.
         horizons: Horizons to train on in hours, relevant for feature engineering.
+        ignore_existing_models: If True, all existing models, including, hyperparameters are ignored and defsault values are used. 
 
     Raises:
         InputDataInsufficientError: when input data is insufficient.
@@ -319,8 +321,9 @@ def train_pipeline_step_load_model(
     old_model: Optional[OpenstfRegressor]
 
     if not ignore_existing_models:
+        model_run_id = pj.get("model_run_id", None)
         try:
-            old_model, model_specs = serializer.load_model(experiment_name=str(pj.id))
+            old_model, model_specs = serializer.load_model(str(pj.id), model_run_id=model_run_id)
             old_model_age = old_model.age  # Age attribute is openstef specific
             return old_model, model_specs, old_model_age
         except (AttributeError, FileNotFoundError, LookupError):

{openstef-3.4.77 → openstef-3.4.79}/openstef.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: openstef
-Version: 3.4.77
+Version: 3.4.79
 Summary: Open short term energy forecaster
 Home-page: https://github.com/OpenSTEF/openstef
 Author: Alliander N.V

{openstef-3.4.77 → openstef-3.4.79}/setup.py

@@ -33,7 +33,7 @@ def read_long_description_from_readme():
 
 setup(
     name="openstef",
-    version="3.4.77",
+    version="3.4.79",
     packages=find_packages(include=["openstef", "openstef.*"]),
     description="Open short term energy forecaster",
     long_description=read_long_description_from_readme(),