[python] dalex v1.6.1

.gitignore

@@ -1,3 +1,6 @@
+.vscode/settings.json
+**.DS_Store
+
 .Rproj.user
 .Rhistory
 .RData

python/dalex/NEWS.md

@@ -1,10 +1,14 @@
 ## Changelog
 
-### development
 
-* Fix an error occuring in `predict_profile()` when a DataFrame has MultiIndex in `pandas >= 1.3.0` ([#550](https://github.com/ModelOriented/DALEX/pull/550))
+### v1.6.1 (2024-02-28)
+
+* Added `keras.src.models.sequential.Sequential` to classes with a known `predict_function`; it should fix changes in `keras==3.0.0` and `tensorflow==2.16.0`
+* Turn off `verbose` in the predict method of tensorflow/keras models that [changed](https://stackoverflow.com/a/73244830) in `tensorflow>=2.9.0`
+* Fix an error occuring in `predict_profile()` when a DataFrame has MultiIndex in `pandas>=1.3.0` ([#550](https://github.com/ModelOriented/DALEX/pull/550))
 * Fix gaussian `norm()` calculation in `model_profile()` from `pi*sqrt(2)` to `sqrt(2*pi)`
-* Potential fix for a warning (future error) between `prepare_numerical_categorical()` and `prepare_x()` with `pandas == 2.1.0`
+* Fix a warning (future error) between `prepare_numerical_categorical()` and `prepare_x()` with `pandas==2.1.0`
+* Fix a warning (future error) concerning the default value of `numeric_only` in `pandas.DataFrame.corr()` in `dalex.aspect.calculate_assoc_matrix()`
 
 
 ### v1.6.0 (2023-02-16) 

python/dalex/dalex/__init__.py

@@ -9,7 +9,7 @@
 from .aspect import Aspect
 
 
-__version__ = '1.6.0.9000'
+__version__ = '1.6.1'
 
 __all__ = [
   "Arena",

python/dalex/dalex/_explainer/checks.py

@@ -1,8 +1,8 @@
 # check functions for Explainer.__init__
 import numpy as np
 import pandas as pd
+import warnings
 from copy import deepcopy
-from warnings import warn
 
 from .helper import verbose_cat, is_y_in_data, get_model_info
 from .yhat import get_predict_function_and_model_type
@@ -199,7 +199,9 @@ def check_predict_function_and_model_type(predict_function, model_type,
         # check if predict_function accepts arrays
         try:
             data_values = data.values[[0]]
-            predict_function(model, data_values)
+            with warnings.catch_warnings():
+                warnings.simplefilter('ignore') # ignore warnings about feature names in scikit-learn
+                predict_function(model, data_values)
             model_info_['arrays_accepted'] = True
             verbose_cat("  -> predict function  : Accepts pandas.DataFrame and numpy.ndarray.",
                         verbose=verbose)
@@ -219,11 +221,11 @@ def check_predict_function_and_model_type(predict_function, model_type,
                 # verbose_cat("  -> predicted values  : the predict_function returns an error when executed \n",
                 #             verbose=verbose)
 
-                warn("\n  -> predicted values  : 'predict_function' returns an Error when executed: \n" +
+                warnings.warn("\n  -> predicted values  : 'predict_function' returns an Error when executed: \n" +
                      str(error), stacklevel=2)
 
             if not isinstance(y_hat, np.ndarray) or y_hat.shape != (data.shape[0], ):
-                warn("\n  -> predicted values  : 'predict_function' must return numpy.ndarray (1d)", stacklevel=2)
+                warnings.warn("\n  -> predicted values  : 'predict_function' must return numpy.ndarray (1d)", stacklevel=2)
 
     if model_type is None:
         # model_type not specified

python/dalex/dalex/_explainer/yhat.py

@@ -17,7 +17,8 @@ def yhat_xgboost(m, d):
 
 def get_tf_yhat(model):
     if not (str(type(model)).startswith("<class 'tensorflow.python.keras.engine") or
-             str(type(model)).startswith("<class 'keras.engine")):
+             str(type(model)).startswith("<class 'keras.engine") or
+             str(type(model)).startswith("<class 'keras.src.models")):
         return None
 
     if model.output_shape[1] == 1:
@@ -30,11 +31,11 @@ def get_tf_yhat(model):
 
 
 def yhat_tf_regression(m, d):
-    return m.predict(np.array(d)).reshape(-1, )
+    return m.predict(np.array(d), verbose=0).reshape(-1, )
 
 
 def yhat_tf_classification(m, d):
-    return m.predict(np.array(d))[:, 1]
+    return m.predict(np.array(d), verbose=0)[:, 1]
 
 
 def get_h2o_yhat(model):
@@ -80,6 +81,7 @@ def get_predict_function_and_model_type(model, model_class):
         "keras.engine.sequential.Sequential": prep_tf,
         "keras.engine.training.Model": prep_tf,
         "keras.engine.functional.Functional": prep_tf,
+        "keras.src.models.sequential.Sequential": prep_tf,
         "h2o.estimators.coxph.H2OCoxProportionalHazardsEstimator": prep_h2o,
         "h2o.estimators.deeplearning.H2ODeepLearningEstimator": prep_h2o,
         "h2o.estimators.gam.H2OGeneralizedAdditiveEstimator": prep_h2o,

python/dalex/dalex/aspect/utils.py

@@ -53,7 +53,7 @@ def calculate_cat_num_assoc_matrix(data, categorical_variables, numerical_variab
 
 
 def calculate_assoc_matrix(data, corr_method):
-    corr_matrix = abs(data.corr(corr_method)) # get submatrix for only numerical variables
+    corr_matrix = abs(data.corr(corr_method, numeric_only=True)) # get submatrix for only numerical variables
     numerical_variables = corr_matrix.columns 
     categorical_variables = list(set(data.columns) - set(numerical_variables))
     n = len(data)

python/dalex/test/test_ceteris_paribus.py

@@ -90,15 +90,15 @@ def test_single_variable_profile(self):
 
     def test_calculate_variable_profile(self):
         splits = utils.calculate_variable_split(self.X, ['age', 'gender'], 121)
-        vp = utils.calculate_variable_profile(self.exp.predict_function, self.exp.model, self.X.iloc[[0], :], splits, 1)
+        vp = utils.calculate_variable_profile(self.exp.predict_function, self.exp.model, self.X.iloc[[0], :], splits, 1, verbose=False)
         self.assertIsInstance(vp, pd.DataFrame)
 
         splits = utils.calculate_variable_split(self.X, ['gender'], 5)
-        vp = utils.calculate_variable_profile(self.exp.predict_function, self.exp.model, self.X.iloc[[0], :], splits, 1)
+        vp = utils.calculate_variable_profile(self.exp.predict_function, self.exp.model, self.X.iloc[[0], :], splits, 1, verbose=False)
         self.assertIsInstance(vp, pd.DataFrame)
 
         splits = utils.calculate_variable_split(self.X, self.X.columns, 15)
-        vp = utils.calculate_variable_profile(self.exp.predict_function, self.exp.model, self.X.iloc[[0], :], splits, 2)
+        vp = utils.calculate_variable_profile(self.exp.predict_function, self.exp.model, self.X.iloc[[0], :], splits, 2, verbose=False)
         self.assertIsInstance(vp, pd.DataFrame)
 
     def test_calculate_ceteris_paribus(self):

python/dalex/test/test_fairness.py

@@ -1,9 +1,9 @@
 import unittest
-from copy import deepcopy
+from copy import copy, deepcopy
+import warnings
 
 import numpy as np
 import pandas as pd
-from copy import copy
 from plotly.graph_objs import Figure
 from sklearn.compose import ColumnTransformer
 from sklearn.pipeline import Pipeline
@@ -61,7 +61,7 @@ def setUp(self):
         reg = DecisionTreeRegressor()
         reg.fit(data2, target)
 
-        self.exp_reg = dx.Explainer(reg, data2, target)
+        self.exp_reg = dx.Explainer(reg, data2, target, verbose=False)
         self.mgf_reg = self.exp_reg.model_fairness(self.protected_reg, 'A')
 
     def test_fairness_check(self):
@@ -160,7 +160,9 @@ def test_calculate_ratio(self):
         b = list(scf_metrics.subgroup_confusion_matrix_metrics.get('b').values())
         a = list(scf_metrics.subgroup_confusion_matrix_metrics.get('a').values())
 
-        ratio = np.array(b) / np.array(a)
+        with warnings.catch_warnings():
+            warnings.simplefilter('ignore') # ignore warning about division by zero
+            ratio = np.array(b) / np.array(a)
         ratio[np.isinf(ratio)] = np.nan
         ratio[ratio == 0] = np.nan
 

python/dalex/test/test_model_diagnostics.py

@@ -76,8 +76,8 @@ def test_plot(self):
         self.assertIsInstance(case3, dx.model_explanations.ResidualDiagnostics)
 
         fig1 = case1.plot(title="test1", variable="fare", N=1000, show=False)
-        fig2 = case2.plot(case3, variable="sibsp", yvariable="abs_residuals",  N=None, show=False)
-        fig3 = case2.plot(smooth=False, line_width=6, marker_size=1, variable="age", show=False)
+        fig2 = case2.plot(case3, variable="age", yvariable="abs_residuals",  N=None, show=False)
+        fig3 = case2.plot(smooth=False, line_width=6, marker_size=1, variable="sibsp", show=False)
 
         self.assertIsInstance(fig1, Figure)
         self.assertIsInstance(fig2, Figure)