Less freezing time in multiprocessing with large arrays

tests.py

@@ -263,6 +263,7 @@ def isclose(a, b, rel_tol=1e-09, abs_tol=0.0):
 
     assert isclose(known_score, score)
 
+
 def test_score_3():
     """Assert that the TPOTRegressor score function outputs a known score for a fix pipeline"""
 
@@ -293,6 +294,7 @@ def isclose(a, b, rel_tol=1e-09, abs_tol=0.0):
 
     assert isclose(known_score, score)
 
+
 def test_sample_weight_func():
     """Assert that the TPOTRegressor score function outputs a known score for a fixed pipeline with sample weights"""
 
@@ -342,6 +344,7 @@ def isclose(a, b, rel_tol=1e-09, abs_tol=0.0):
     assert not np.allclose(cv_score1, cv_score_weight)
     assert isclose(known_score, score)
 
+
 def test_predict():
     """Assert that the TPOT predict function raises a RuntimeError when no optimized pipeline exists"""
 
@@ -369,6 +372,7 @@ def test_predict_2():
 
     assert result.shape == (testing_features.shape[0],)
 
+
 def test_predict_proba():
     """Assert that the TPOT predict_proba function returns a numpy matrix of shape (num_testing_rows, num_testing_classes)"""
 
@@ -410,6 +414,7 @@ def test_predict_proba2():
     except Exception:
         assert False
 
+
 def test_warm_start():
     """Assert that the TPOT warm_start flag stores the pop and pareto_front from the first run"""
     tpot_obj = TPOTClassifier(random_state=42, population_size=1, offspring_size=2, generations=1, verbosity=0, warm_start=True)
@@ -582,6 +587,34 @@ def test_generate_import_code():
 """
     assert expected_code == generate_import_code(pipeline, tpot_obj.operators)
 
+
+def test_PolynomialFeatures_exception():
+    """Assert"""
+    tpot_obj = TPOTClassifier()
+    tpot_obj._pbar = tqdm(total=1, disable=True)
+    # pipeline with one PolynomialFeatures operator
+    pipeline_string_1 = ('LogisticRegression(PolynomialFeatures'
+    '(input_matrix, PolynomialFeatures__degree=2, PolynomialFeatures__include_bias=DEFAULT, '
+    'PolynomialFeatures__interaction_only=False), LogisticRegression__C=10.0, '
+    'LogisticRegression__dual=DEFAULT, LogisticRegression__penalty=DEFAULT)')
+
+    # pipeline with two PolynomialFeatures operator
+    pipeline_string_2 = ('LogisticRegression(PolynomialFeatures'
+    '(PolynomialFeatures(input_matrix, PolynomialFeatures__degree=2, '
+    'PolynomialFeatures__include_bias=DEFAULT, PolynomialFeatures__interaction_only=False), '
+    'PolynomialFeatures__degree=2, PolynomialFeatures__include_bias=DEFAULT, '
+    'PolynomialFeatures__interaction_only=False), LogisticRegression__C=10.0, '
+    'LogisticRegression__dual=DEFAULT, LogisticRegression__penalty=DEFAULT)')
+
+    # make a list for _evaluate_individuals
+    pipelines = []
+    pipelines.append(creator.Individual.from_string(pipeline_string_1, tpot_obj._pset))
+    pipelines.append(creator.Individual.from_string(pipeline_string_2, tpot_obj._pset))
+    fitness_scores = tpot_obj._evaluate_individuals(pipelines, training_features, training_classes)
+    known_scores = [(2, 0.98068077235290885), (5000.0, -float('inf'))]
+    assert np.allclose(known_scores, fitness_scores)
+
+
 def test_mutNodeReplacement():
     """Assert that mutNodeReplacement() returns the correct type of mutation node in a fixed pipeline"""
     tpot_obj = TPOTClassifier()
@@ -649,6 +682,7 @@ def test_export_pipeline():
 """
     assert expected_code == export_pipeline(pipeline, tpot_obj.operators, tpot_obj._pset)
 
+
 def test_export_pipeline_2():
     """Assert that exported_pipeline() generated a compile source file as expected given a fixed simple pipeline (only one classifier)"""
     tpot_obj = TPOTClassifier()
@@ -673,6 +707,7 @@ def test_export_pipeline_2():
 """
     assert expected_code == export_pipeline(pipeline, tpot_obj.operators, tpot_obj._pset)
 
+
 def test_export_pipeline_3():
     """Assert that exported_pipeline() generated a compile source file as expected given a fixed simple pipeline with a preprocessor"""
     tpot_obj = TPOTClassifier()
@@ -704,6 +739,7 @@ def test_export_pipeline_3():
 """
     assert expected_code == export_pipeline(pipeline, tpot_obj.operators, tpot_obj._pset)
 
+
 def test_operator_export():
     """Assert that a TPOT operator can export properly with a function as a parameter to a classifier"""
     export_string = TPOTSelectKBest.export(5)

tpot/base.py

@@ -705,13 +705,13 @@ def _evaluate_individuals(self, individuals, features, classes, sample_weight =
             # This is a fairly hacky way to prevent TPOT from getting stuck on bad pipelines and should be improved in a future release
             individual = individuals[indidx]
             individual_str = str(individual)
-            if individual_str.count('PolynomialFeatures') > 1:
+            sklearn_pipeline_str = generate_pipeline_code(expr_to_tree(individual, self._pset), self.operators)
+            if sklearn_pipeline_str.count('PolynomialFeatures') > 1:
                 if self.verbosity > 2:
                     self._pbar.write('Invalid pipeline encountered. Skipping its evaluation.')
                 fitnesses_dict[indidx] = (5000., -float('inf'))
                 if not self._pbar.disable:
                     self._pbar.update(1)
-
             # Check if the individual was evaluated before
             elif individual_str in self._evaluated_individuals:
                 # Get fitness score from previous evaluation
@@ -726,6 +726,7 @@ def _evaluate_individuals(self, individuals, features, classes, sample_weight =
                     # Transform the tree expression into an sklearn pipeline
                     sklearn_pipeline = self._toolbox.compile(expr=individual)
 
+
                     # Fix random state when the operator allows and build sample weight dictionary
                     self._set_param_recursive(sklearn_pipeline.steps, 'random_state', 42)