Fix regression preprocessing

cases/credit_scoring/credit_scoring_problem_multiobj.py

@@ -7,11 +7,14 @@
 from cases.credit_scoring.credit_scoring_problem import get_scoring_data
 from fedot.core.composer.composer_builder import ComposerBuilder
 from fedot.core.data.data import InputData
-from fedot.core.optimisers.gp_comp.gp_optimizer import GeneticSchemeTypesEnum
 from fedot.core.optimisers.gp_comp.gp_params import GPGraphOptimizerParameters
+from fedot.core.optimisers.gp_comp.operators.inheritance import GeneticSchemeTypesEnum
 from fedot.core.optimisers.gp_comp.operators.selection import SelectionTypesEnum
 from fedot.core.optimisers.gp_comp.pipeline_composer_requirements import PipelineComposerRequirements
+from fedot.core.pipelines.node import PrimaryNode
 from fedot.core.pipelines.pipeline import Pipeline
+from fedot.core.pipelines.tuning.sequential import SequentialTuner
+from fedot.core.pipelines.tuning.tuner_builder import TunerBuilder
 from fedot.core.repository.operation_types_repository import get_operations_for_task
 from fedot.core.repository.quality_metrics_repository import ClassificationMetricsEnum, ComplexityMetricsEnum
 from fedot.core.repository.tasks import Task, TaskTypesEnum
@@ -57,7 +60,8 @@ def run_credit_scoring_problem(train_file_path, test_file_path,
     composer_requirements = PipelineComposerRequirements(
         primary=available_model_types,
         secondary=available_model_types,
-        timeout=timeout
+        timeout=timeout,
+        num_of_generations=20
     )
     params = GPGraphOptimizerParameters(
         selection_types=[SelectionTypesEnum.spea2],
@@ -80,10 +84,18 @@ def run_credit_scoring_problem(train_file_path, test_file_path,
         results_visualization(composed_pipelines=pipelines_evo_composed, history=composer.history)
 
     pipelines_roc_auc = []
+
     for pipeline_num, pipeline_evo_composed in enumerate(pipelines_evo_composed):
 
-        pipeline_evo_composed.fine_tune_primary_nodes(input_data=dataset_to_compose,
-                                                      iterations=50)
+        tuner = TunerBuilder(task)\
+            .with_tuner(SequentialTuner)\
+            .with_iterations(50)\
+            .with_metric(metrics[0])\
+            .build(dataset_to_compose)
+        nodes = pipeline_evo_composed.nodes
+        for node_index, node in enumerate(nodes):
+            if isinstance(node, PrimaryNode):
+                pipeline_evo_composed = tuner.tune_node(pipeline_evo_composed, node_index)
 
         pipeline_evo_composed.fit(input_data=dataset_to_compose)
 

cases/river_levels_prediction/river_level_case_manual.py

@@ -44,7 +44,7 @@ def run_river_experiment(file_path, pipeline, iterations=20, tuner=None,
         current_pipeline = copy(pipeline)
 
         # Fit it
-        current_pipeline.fit_from_scratch(train_input)
+        current_pipeline.fit(train_input)
 
         # Predict
         predicted_values = current_pipeline.predict(predict_input)
@@ -68,6 +68,9 @@ def run_river_experiment(file_path, pipeline, iterations=20, tuner=None,
                 .build(train_input)
             tuned_pipeline = pipeline_tuner.tune(current_pipeline)
 
+            # Fit it
+            tuned_pipeline.fit(train_input)
+
             # Predict
             predicted_values_tuned = tuned_pipeline.predict(predict_input)
             preds_tuned = predicted_values_tuned.predict

examples/advanced/decompose/classification_refinement_example.py

@@ -74,6 +74,9 @@ def run_refinement_scoring_example(train_path, test_path, with_tuning=False):
         no_decompose_c = tuner.tune(no_decompose_c)
         decompose_c = tuner.tune(decompose_c)
 
+        no_decompose_c.fit(test_dataset)
+        decompose_c.fit(test_dataset)
+
         display_roc_auc(no_decompose_c, test_dataset, 'Non decomposition pipeline after tuning')
         display_roc_auc(decompose_c, test_dataset, 'With decomposition pipeline after tuning')
 

examples/advanced/decompose/regression_refinement_example.py

@@ -95,10 +95,6 @@ def run_river_experiment(file_path, with_tuning=False):
     r_pipeline = get_refinement_pipeline()
     non_pipeline = get_non_refinement_pipeline()
 
-    # Fit it
-    r_pipeline.fit(train_input)
-    non_pipeline.fit(train_input)
-
     if with_tuning:
         tuner = TunerBuilder(task)\
             .with_tuner(PipelineTuner)\
@@ -108,6 +104,10 @@ def run_river_experiment(file_path, with_tuning=False):
         r_pipeline = tuner.tune(r_pipeline)
         non_pipeline = tuner.tune(non_pipeline)
 
+    # Fit it
+    r_pipeline.fit(train_input)
+    non_pipeline.fit(train_input)
+
     # Predict
     predicted_values = r_pipeline.predict(predict_input)
     r_preds = predicted_values.predict

examples/simple/classification/classification_with_tuning.py

@@ -109,7 +109,7 @@ def run_classification_tuning_experiment(pipeline, tuner=None):
                                   data_type=DataTypesEnum.table)
 
         # Fit it
-        pipeline.fit_from_scratch(train_input)
+        pipeline.fit(train_input)
 
         # Predict
         predicted_labels = pipeline.predict(predict_input)
@@ -126,6 +126,9 @@ def run_classification_tuning_experiment(pipeline, tuner=None):
                 .build(train_input)
             tuned_pipeline = pipeline_tuner.tune(pipeline)
 
+            # Fit it
+            pipeline.fit(train_input)
+
             # Predict
             print('predict')
             predicted_values_tuned = tuned_pipeline.predict(predict_input)

examples/simple/classification/resample_example.py

@@ -74,7 +74,7 @@ def run_resample_example(path_to_data=None, tune=False):
 
     print('Begin fit Pipeline with balancing')
     # pipeline.fit(train_input)
-    pipeline.fit_from_scratch(train_input)
+    pipeline.fit(train_input)
 
     # Predict
     predict_labels = pipeline.predict(predict_input)
@@ -91,7 +91,8 @@ def run_resample_example(path_to_data=None, tune=False):
             .with_timeout(timedelta(minutes=1))\
             .build(train_input)
         tuned_pipeline = tuner.tune(pipeline)
-
+        # Fit
+        pipeline.fit(train_input)
         # Predict
         predicted_values_tuned = tuned_pipeline.predict(predict_input)
         preds_tuned = predicted_values_tuned.predict

examples/simple/regression/regression_with_tuning.py

@@ -89,7 +89,7 @@ def run_experiment(pipeline, tuner):
                                   data_type=DataTypesEnum.table)
 
         # Fit it
-        pipeline.fit_from_scratch(train_input)
+        pipeline.fit(train_input)
 
         # Predict
         predicted_values = pipeline.predict(predict_input)
@@ -108,6 +108,9 @@ def run_experiment(pipeline, tuner):
                 .build(train_input)
             tuned_pipeline = pipeline_tuner.tune(pipeline)
 
+            # Fit it
+            tuned_pipeline.fit(train_input)
+
             # Predict
             predicted_values_tuned = tuned_pipeline.predict(predict_input)
             preds_tuned = predicted_values_tuned.predict
@@ -117,6 +120,8 @@ def run_experiment(pipeline, tuner):
             print('Obtained metrics after tuning:')
             print(f'MAE - {mae_value:.4f}\n')
 
+        pipeline.unfit()
+
 
 # Script for testing is pipeline can process different datasets for regression task
 if __name__ == '__main__':

examples/simple/time_series_forecasting/tuning_pipelines.py

@@ -77,7 +77,7 @@ def run_experiment(dataset: str, pipeline: Pipeline, len_forecast=250, tuning=Tr
             .with_iterations(100) \
             .build(train_data)
         pipeline = tuner.tune(pipeline)
-
+        pipeline.fit(train_data)
         prediction_after = pipeline.predict(test_data)
         predict_after = np.ravel(np.array(prediction_after.predict))
 

fedot/core/optimisers/objective/data_objective_eval.py

@@ -41,16 +41,14 @@ def __init__(self,
                  validation_blocks: Optional[int] = None,
                  pipelines_cache: Optional[OperationsCache] = None,
                  preprocessing_cache: Optional[PreprocessingCache] = None,
-                 eval_n_jobs: int = 1,
-                 do_unfit: bool = True):
+                 eval_n_jobs: int = 1):
         super().__init__(objective, eval_n_jobs=eval_n_jobs)
         self._data_producer = data_producer
         self._time_constraint = time_constraint
         self._validation_blocks = validation_blocks
         self._pipelines_cache = pipelines_cache
         self._preprocessing_cache = preprocessing_cache
         self._log = default_log(self)
-        self._do_unfit = do_unfit
 
     def evaluate(self, graph: Pipeline) -> Fitness:
         # Seems like a workaround for situation when logger is lost
@@ -83,8 +81,7 @@ def evaluate(self, graph: Pipeline) -> Fitness:
                     raise ValueError(f'Fitness {evaluated_fitness} is not valid')
                 else:
                     continue
-            if self._do_unfit:
-                graph.unfit()
+            graph.unfit()
         if folds_metrics:
             folds_metrics = tuple(np.mean(folds_metrics, axis=0))  # averages for each metric over folds
             self._log.debug(f'Pipeline {graph_id} with evaluated metrics: {folds_metrics}')

fedot/core/pipelines/tuning/tuner_builder.py

@@ -83,7 +83,7 @@ def build(self, data: InputData) -> HyperoptTuner:
         data_producer = DataSourceSplitter(self.cv_folds, self.validation_blocks).build(data)
         objective_evaluate = PipelineObjectiveEvaluate(objective, data_producer,
                                                        validation_blocks=self.validation_blocks,
-                                                       do_unfit=False, time_constraint=self.eval_time_constraint)
+                                                       time_constraint=self.eval_time_constraint)
         tuner = self.tuner_class(objective_evaluate=objective_evaluate,
                                  iterations=self.iterations,
                                  early_stopping_rounds=self.early_stopping_rounds,

fedot/core/pipelines/tuning/tuner_interface.py

@@ -120,4 +120,4 @@ def final_check(self, tuned_pipeline: Pipeline):
         else:
             self.log.info(f'{prefix_init_phrase} {abs(self.obtained_metric):.3f} '
                           f'worse than initial (+ 5% deviation) {abs(init_metric):.3f}')
-            return self.init_pipeline
+            return self.init_pipeline