make featuretools optional

hypernets/experiment/compete.py

@@ -1894,7 +1894,8 @@ def __init__(self, hyper_model, X_train, y_train, X_eval=None, y_eval=None, X_te
 
  if feature_generation:
  if 'FeatureGenerationTransformer' not in tb.transformers.keys():
- raise NotImplementedError('feature_generation is not supported for your data')
+ raise ValueError('feature_generation is not supported for your data, '
+ 'or "featuretools" is not installed.')
 
  if data_cleaner_args is None:
  data_cleaner_args = {}

hypernets/tabular/dask_ex/_feature_generators.py

@@ -4,6 +4,9 @@
 """
 
 from ..feature_generators import FeatureGenerationTransformer
+from ..feature_generators import is_feature_generator_ready as _is_feature_generator_ready
+
+is_feature_generator_ready = _is_feature_generator_ready
 
 
 class DaskFeatureGenerationTransformer(FeatureGenerationTransformer):

hypernets/tabular/dask_ex/_toolbox.py

@@ -820,9 +820,12 @@ def compute_sample_weight(y):
 
  # TfidfEncoder=sk_ex.TfidfEncoder,
  # DatetimeEncoder=sk_ex.DatetimeEncoder,
- FeatureGenerationTransformer=_feature_generators.DaskFeatureGenerationTransformer,
+ # FeatureGenerationTransformer=_feature_generators.DaskFeatureGenerationTransformer,
  FeatureImportancesSelectionTransformer=sk_ex.FeatureImportancesSelectionTransformer,
 )
 
+if _feature_generators.is_feature_generator_ready:
+ _predefined_transformers['FeatureGenerationTransformer'] = _feature_generators.FeatureGenerationTransformer
+
 for name, tf in _predefined_transformers.items():
  register_transformer(tf, name=name, dtypes=dd.DataFrame)

hypernets/tabular/feature_generators/__init__.py

@@ -2,5 +2,22 @@
 """
 
 """
-from ._primitives import CrossCategorical, GeoHashPrimitive, DaskCompatibleHaversine, TfidfPrimitive
-from ._transformers import FeatureGenerationTransformer, is_geohash_installed
+# from ._primitives import CrossCategorical, GeoHashPrimitive, DaskCompatibleHaversine, TfidfPrimitive
+# from ._transformers import FeatureGenerationTransformer, is_geohash_installed
+
+try:
+ from ._transformers import FeatureGenerationTransformer, is_geohash_installed
+
+ is_feature_generator_ready = True
+except ImportError as e:
+ _msg = f'{e}, install featuretools and try again'
+
+ is_geohash_installed = False
+ is_feature_generator_ready = False
+
+ from sklearn.base import BaseEstimator as _BaseEstimator
+
+
+ class FeatureGenerationTransformer(_BaseEstimator):
+ def __init__(self, *args, **kwargs):
+ raise ImportError(_msg)

hypernets/tabular/toolbox.py

@@ -724,9 +724,12 @@ def greedy_ensemble(cls, task, estimators, need_fit=False, n_folds=5, method='so
  # TfidfEncoder=sk_ex.TfidfEncoder,
  # DatetimeEncoder=sk_ex.DatetimeEncoder,
 
- FeatureGenerationTransformer=feature_generators_.FeatureGenerationTransformer,
+ # FeatureGenerationTransformer=feature_generators_.FeatureGenerationTransformer,
 )
 
+if feature_generators_.is_feature_generator_ready:
+ _predefined_transformers['FeatureGenerationTransformer'] = feature_generators_.FeatureGenerationTransformer
+
 for name, tf in _predefined_transformers.items():
  register_transformer(tf, name=name, dtypes=pd.DataFrame)
 

hypernets/tests/experiment/compete_experiment_test.py

@@ -1,17 +1,18 @@
 from datetime import datetime
 
 import numpy as np
-
+import pytest
 from sklearn.preprocessing import LabelEncoder
 
 from hypernets.core import SummaryCallback
 from hypernets.core.objective import Objective
 from hypernets.examples.plain_model import PlainModel, PlainSearchSpace
 from hypernets.experiment import CompeteExperiment
-from hypernets.model.objectives import ElapsedObjective, PredictionObjective
+from hypernets.model.objectives import PredictionObjective
 from hypernets.searchers.nsga_searcher import NSGAIISearcher
 from hypernets.tabular import get_tool_box
 from hypernets.tabular.datasets import dsutils
+from hypernets.tabular.feature_generators import is_feature_generator_ready
 from hypernets.tabular.sklearn_ex import MultiLabelEncoder
 from hypernets.tests.model.plain_model_test import create_plain_model
 from hypernets.tests.tabular.tb_dask import if_dask_ready, is_dask_installed, setup_dask
@@ -150,6 +151,7 @@ def test_without_cv():
  experiment_with_bank_data(dict(cv=False), {})
 
 
+@pytest.mark.skipif(not is_feature_generator_ready, reason='feature_generator is not ready')
 def test_with_feature_generation():
  experiment_with_movie_lens(dict(feature_generation=True,
  feature_generation_text_cols=['title']), {})
@@ -199,7 +201,8 @@ def test_with_pi():
  feature_reselection_threshold=0.0001), {})
 
 
-def test_with_feature_generator():
+@pytest.mark.skipif(not is_feature_generator_ready, reason='feature_generator is not ready')
+def test_with_feature_generation_and_selection():
  experiment_with_movie_lens(dict(feature_generation=True, feature_selection=True,
  feature_generation_text_cols=['title']), {})
 
@@ -223,6 +226,7 @@ def test_with_cv_ensemble_dask():
 
 
 @if_dask_ready
+@pytest.mark.skipif(not is_feature_generator_ready, reason='feature_generator is not ready')
 def test_with_feature_generator_dask():
  experiment_with_movie_lens(dict(feature_generation=True, feature_selection=True,
  feature_generation_text_cols=['title']), {}, with_dask=True)

hypernets/tests/experiment/extractor_test.py

@@ -1,7 +1,8 @@
 from hypernets.tests.experiment import experiment_factory
 from hypernets.experiment.compete import DataCleanStep, DriftDetectStep
 from hypernets.experiment import ExperimentExtractor, StepMeta
-import time
+import pytest
+from hypernets.tabular.feature_generators import is_feature_generator_ready
 
 
 def _run_experiment(creator):
@@ -73,6 +74,7 @@ def test_multicollinearity_detect_extractor():
  assert unselected_features['INDUS']['reserved'] == 'CRIM'
 
 
+@pytest.mark.skipif(not is_feature_generator_ready, reason='feature_generator is not ready')
 def test_feature_generation_extractor():
  exp_data, estimator = _run_experiment(experiment_factory.create_feature_generation_experiment)
  fg_step = exp_data.steps[2]

hypernets/tests/experiment/make_experiment_test.py

@@ -9,6 +9,7 @@
 from hypernets.experiment.compete import StepNames
 from hypernets.tabular import get_tool_box
 from hypernets.tabular.datasets import dsutils
+from hypernets.tabular.feature_generators import is_feature_generator_ready
 from hypernets.tabular.sklearn_ex import MultiLabelEncoder
 from hypernets.utils import common as common_util
 from hypernets.searchers.nsga_searcher import NSGAIISearcher
@@ -67,7 +68,7 @@ def test_experiment_with_blood_full_features():
 
  experiment = make_experiment(PlainModel, df, target=target, search_space=PlainSearchSpace(),
  test_data=df_test,
- feature_generation=True,
+ feature_generation=is_feature_generator_ready,
  collinearity_detection=True,
  drift_detection=True,
  feature_selection=True,

hypernets/tests/tabular/feature_generator_test.py

@@ -5,7 +5,6 @@
 import math
 from datetime import datetime
 
-import featuretools as ft
 import numpy as np
 import pandas as pd
 import pytest
@@ -17,7 +16,8 @@
 from hypernets.tabular.column_selector import column_object_category_bool, column_number_exclude_timedelta
 from hypernets.tabular.dataframe_mapper import DataFrameMapper
 from hypernets.tabular.datasets import dsutils
-from hypernets.tabular.feature_generators import FeatureGenerationTransformer, is_geohash_installed
+from hypernets.tabular.feature_generators import FeatureGenerationTransformer, is_geohash_installed, \
+ is_feature_generator_ready
 from hypernets.tabular.sklearn_ex import FeatureSelectionTransformer
 from hypernets.utils import logging
 
@@ -38,6 +38,7 @@ def general_preprocessor():
  return preprocessor
 
 
+@pytest.mark.skipif(not is_feature_generator_ready, reason='feature_generator is not ready')
 class Test_FeatureGenerator():
  def test_char_add(self):
  x1 = ['1', '2']
@@ -46,6 +47,7 @@ def test_char_add(self):
  assert list(x3) == ['1c', '2d']
 
  def test_ft_primitives(self):
+ import featuretools as ft
  tps = ft.primitives.get_transform_primitives()
  assert tps
 

hypernets/tests/tabular/tb_dask/feature_generator_dask_test.py

@@ -11,7 +11,7 @@
 
 from hypernets.tabular import get_tool_box
 from hypernets.tabular.datasets import dsutils
-from hypernets.tabular.feature_generators import is_geohash_installed
+from hypernets.tabular.feature_generators import is_geohash_installed, is_feature_generator_ready
 from hypernets.utils import logging
 from . import if_dask_ready, is_dask_installed, setup_dask
 
@@ -24,6 +24,7 @@
 
 
 @if_dask_ready
+@pytest.mark.skipif(not is_feature_generator_ready, reason='feature_generator is not ready')
 class TestFeatureGeneratorWithDask:
  @classmethod
  def setup_class(cls):