erroranalysis 0.5.4__tar.gz → 0.5.5__tar.gz

{erroranalysis-0.5.4/erroranalysis.egg-info → erroranalysis-0.5.5}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: erroranalysis
-Version: 0.5.4
+Version: 0.5.5
 Summary: Core error analysis APIs
 Home-page: https://github.com/microsoft/responsible-ai-widgets
 Author: Roman Lutz, Ilya Matiach, Ke Xu
@@ -16,7 +16,7 @@ Classifier: Development Status :: 3 - Alpha
 Requires-Python: >=3.7
 Description-Content-Type: text/markdown
 License-File: LICENSE
-Requires-Dist: numpy>=1.17.2
+Requires-Dist: numpy<2.0.0,>=1.17.2
 Requires-Dist: pandas<2.0.0,>=0.25.1
 Requires-Dist: scipy>=1.4.1
 Requires-Dist: scikit-learn>=0.22.1

{erroranalysis-0.5.4 → erroranalysis-0.5.5}/erroranalysis/_internal/surrogate_error_tree.py

@@ -53,6 +53,7 @@ DEFAULT_NUM_LEAVES = 31
 DEFAULT_MIN_CHILD_SAMPLES = 20
 CACHED_SUBTREE_FEATURES = 'cache_subtree_features'
 LEAF_VALUE = 'leaf_value'
+OBJECT = 'object'
 PREDICTION = 'prediction'
 RAW_PREDICTION = 'rawPrediction'
 PROBABILITY = 'probability'
@@ -341,10 +342,20 @@ def get_surrogate_booster_local(filtered_df, analyzer, is_model_analyzer,
         diff = pred_y - true_y
     if not isinstance(diff, np.ndarray):
         diff = np.array(diff)
+
+    # Note: if direct conversion fails, for more complex data scenarios like
+    # object detection we need to convert to an object type for newer versions
+    # of numpy>=1.26.0
     if not isinstance(pred_y, np.ndarray):
-        pred_y = np.array(pred_y)
+        try:
+            pred_y = np.array(pred_y)
+        except ValueError:
+            pred_y = np.array(pred_y, dtype=OBJECT)
     if not isinstance(true_y, np.ndarray):
-        true_y = np.array(true_y)
+        try:
+            true_y = np.array(true_y)
+        except ValueError:
+            true_y = np.array(true_y, dtype=OBJECT)
     if is_pandas:
         input_data = input_data.to_numpy(copy=True)
 

{erroranalysis-0.5.4 → erroranalysis-0.5.5}/erroranalysis/version.py

@@ -4,5 +4,5 @@
 name = 'erroranalysis'
 _major = '0'
 _minor = '5'
-_patch = '4'
+_patch = '5'
 version = '{}.{}.{}'.format(_major, _minor, _patch)

{erroranalysis-0.5.4 → erroranalysis-0.5.5/erroranalysis.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: erroranalysis
-Version: 0.5.4
+Version: 0.5.5
 Summary: Core error analysis APIs
 Home-page: https://github.com/microsoft/responsible-ai-widgets
 Author: Roman Lutz, Ilya Matiach, Ke Xu
@@ -16,7 +16,7 @@ Classifier: Development Status :: 3 - Alpha
 Requires-Python: >=3.7
 Description-Content-Type: text/markdown
 License-File: LICENSE
-Requires-Dist: numpy>=1.17.2
+Requires-Dist: numpy<2.0.0,>=1.17.2
 Requires-Dist: pandas<2.0.0,>=0.25.1
 Requires-Dist: scipy>=1.4.1
 Requires-Dist: scikit-learn>=0.22.1

{erroranalysis-0.5.4 → erroranalysis-0.5.5}/erroranalysis.egg-info/requires.txt

@@ -1,4 +1,4 @@
-numpy>=1.17.2
+numpy<2.0.0,>=1.17.2
 pandas<2.0.0,>=0.25.1
 scipy>=1.4.1
 scikit-learn>=0.22.1

{erroranalysis-0.5.4 → erroranalysis-0.5.5}/requirements.txt

@@ -1,4 +1,4 @@
-numpy>=1.17.2
+numpy>=1.17.2,<2.0.0
 pandas>=0.25.1,<2.0.0
 scipy>=1.4.1
 scikit-learn>=0.22.1

{erroranalysis-0.5.4 → erroranalysis-0.5.5}/tests/test_surrogate_error_tree.py

@@ -9,6 +9,7 @@ import numpy as np
 import pandas as pd
 import pytest
 from common_utils import replicate_dataset
+from ml_wrappers import wrap_model
 
 from erroranalysis._internal.cohort_filter import filter_from_cohort
 from erroranalysis._internal.constants import (ARG, COLUMN, COMPOSITE_FILTERS,
@@ -16,8 +17,9 @@ from erroranalysis._internal.constants import (ARG, COLUMN, COMPOSITE_FILTERS,
                                                OPERATION, PRED_Y, ROW_INDEX,
                                                SPLIT_FEATURE, SPLIT_INDEX,
                                                TRUE_Y, CohortFilterMethods,
-                                               CohortFilterOps, Metrics,
-                                               ModelTask, regression_metrics)
+                                               CohortFilterOps, ImageColumns,
+                                               Metrics, ModelTask,
+                                               regression_metrics)
 from erroranalysis._internal.error_analyzer import (ModelAnalyzer,
                                                     PredictionsAnalyzer)
 from erroranalysis._internal.surrogate_error_tree import (
@@ -28,12 +30,22 @@ from rai_test_utils.datasets.tabular import (
     create_adult_census_data, create_binary_classification_dataset,
     create_cancer_data, create_diabetes_data, create_iris_data,
     create_simple_titanic_data)
+from rai_test_utils.datasets.vision import (
+    get_images, load_fridge_object_detection_dataset)
 from rai_test_utils.models.model_utils import create_models_classification
 from rai_test_utils.models.sklearn import (
     create_kneighbors_classifier, create_sklearn_random_forest_regressor,
     create_titanic_pipeline)
+from rai_test_utils.models.torch import get_object_detection_fridge_model
 from raiutils.exceptions import UserConfigValidationException
 
+try:
+    import torch  # noqa: F401
+    import torchvision  # noqa: F401
+    pytorch_installed = True
+except ImportError:
+    pytorch_installed = False
+
 SIZE = 'size'
 PARENTID = 'parentId'
 ERROR = 'error'
@@ -65,6 +77,71 @@ class DummyInvalidShapeModel:
         return np.zeros((X.shape[0], 1))
 
 
+class SimplifiedWrappedIndexPredictorModel:
+    """Wraps model that uses index to retrieve image data for making
+    predictions. Simplified version of the one in responsibleai-vision
+    package."""
+
+    def __init__(self, model, dataset, image_mode):
+        """Initialize the WrappedIndexPredictorModel.
+
+        :param model: The model to wrap.
+        :type model: object
+        :param dataset: The dataset to use for making predictions.
+        :type dataset: pandas.DataFrame
+        :param image_mode: The mode to open the image in.
+            See pillow documentation for all modes:
+            https://pillow.readthedocs.io/en/stable/handbook/concepts.html
+        :type image_mode: str
+        """
+        self.model = model
+        self.dataset = dataset
+        self.image_mode = image_mode
+        test = get_images(self.dataset, self.image_mode, None)
+        self.predictions = self.model.predict(test)
+        self.predict_proba = self.model.predict_proba(test)
+
+    def index_predictions(self, index, predictions):
+        """Index the predictions.
+
+        :param index: The index to use.
+        :type index: list
+        :param predictions: The predictions to index.
+        :type predictions: list
+        """
+        if not isinstance(index, list):
+            index = list(index)
+        if isinstance(predictions, list):
+            predictions = [predictions[i] for i in index]
+        else:
+            predictions = predictions[index]
+        return predictions
+
+    def predict(self, X):
+        """Predict the class labels for the provided data.
+
+        :param X: Data to predict the labels for.
+        :type X: pandas.DataFrame
+        :return: Predicted class labels.
+        :rtype: list
+        """
+        index = X.index
+        predictions = self.index_predictions(index, self.predictions)
+        return predictions
+
+    def predict_proba(self, X):
+        """Predict the class probabilities for the provided data.
+
+        :param X: Data to predict the probabilities for.
+        :type X: pandas.DataFrame
+        :return: Predicted class probabilities.
+        :rtype: list[list]
+        """
+        index = X.index
+        pred_proba = self.index_predictions(index, self.predict_proba)
+        return pred_proba
+
+
 class TestSurrogateErrorTree(object):
 
     @pytest.mark.parametrize('analyzer_type', [AnalyzerType.MODEL,
@@ -319,6 +396,33 @@ class TestSurrogateErrorTree(object):
             run_error_analyzer(model, X_test, y_test, feature_names,
                                AnalyzerType.MODEL)
 
+    @pytest.mark.skipif(not pytorch_installed,
+                        reason="requires torch/torchvision")
+    def test_surrogate_error_tree_object_detection(self):
+        model_task = ModelTask.OBJECT_DETECTION
+        model = get_object_detection_fridge_model()
+
+        dataset = load_fridge_object_detection_dataset()
+        dataset = dataset.iloc[:3]
+        X_test = dataset[[ImageColumns.IMAGE]]
+        y_test = dataset[[ImageColumns.LABEL]]
+        feature_names = [ImageColumns.IMAGE]
+        image_mode = 'RGB'
+        model = wrap_model(model, X_test, model_task)
+        model = SimplifiedWrappedIndexPredictorModel(model, X_test, image_mode)
+        extracted_feature_names = ["mean_pixel_value", "is_cool_image"]
+        dummy_data = [[141, True], [54, False], [212, True]]
+        ext_dataset = pd.DataFrame(data=dummy_data,
+                                   columns=extracted_feature_names)
+
+        run_error_analyzer(model,
+                           ext_dataset,
+                           y_test,
+                           feature_names,
+                           AnalyzerType.MODEL,
+                           categorical_features=[],
+                           model_task=model_task)
+
 
 def run_error_analyzer(model, X_test, y_test, feature_names,
                        analyzer_type, categorical_features=None,