Interpret Community

0.32.0 verified Fri May 01 auth: no python

Microsoft Interpret Extensions SDK for Python – provides a unified API for model interpretability including SHAP, LIME, and surrogate explainers. Current version: 0.32.0. Release cadence: irregular, ~2-4 per year.

pip install interpret-community

Common errors

error ModuleNotFoundError: No module named 'interpret.ext' ↓

cause Old package versions before 'ext' subpackage was added (or incomplete installation).

fix

Upgrade interpret-community: pip install --upgrade interpret-community

error AttributeError: 'TabularExplainer' object has no attribute 'explain_global' ↓

cause Using an older version where the API was different (pre-v0.20).

fix

Upgrade to v0.20+ and use explain_global() instead of explain().

error ValueError: Expected 2D array, got 1D array instead ↓

cause Passing a single row (1D) to an explainer that expects 2D input (e.g., DataFrame).

fix

Reshape or wrap in a list: explainer.explain_local([X_row]) instead of explainer.explain_local(X_row).

Warnings

breaking In v0.29.0, Python 3.6 support was dropped. Upgrade to Python 3.7+. ↓

fix Use Python 3.7 or higher.

breaking In v0.25.0, the function _get_surrogate_model_replication_measure() was made public as get_surrogate_model_replication_measure(). Private underscore prefix removed. ↓

fix Use get_surrogate_model_replication_measure() instead of _get_surrogate_model_replication_measure().

deprecated The old import path from interpret.blackbox is deprecated. Use interpret.ext.blackbox instead. ↓

fix Use 'from interpret.ext.blackbox import ...'.

gotcha Passing numpy arrays without feature names may cause errors in explainers that expect pandas DataFrames. ↓

fix Wrap input data in a pandas DataFrame with column names when using TabularExplainer.

gotcha When using MimicExplainer, the surrogate model (e.g., DecisionTree) must be passed as a model object, not a string. ↓

fix Instantiate the surrogate model first: from sklearn.tree import DecisionTreeRegressor; MimicExplainer(..., augment_data=True, surrogate_model=DecisionTreeRegressor()).

Imports

TabularExplainer

wrong

from interpret.blackbox import TabularExplainer

correct

from interpret.ext.blackbox import TabularExplainer

Old import path without 'ext' - missing subpackage

MimicExplainer

wrong

from interpret.community import MimicExplainer

correct

from interpret.ext.blackbox import MimicExplainer

MimicExplainer lives in interpret.ext.blackbox, not interpret.community

Quickstart

Basic usage: train a model, create a TabularExplainer, and get global feature importance.

import pandas as pd
from sklearn.datasets import load_iris
from sklearn.ensemble import RandomForestClassifier
from interpret.ext.blackbox import TabularExplainer

iris = load_iris()
X = pd.DataFrame(iris.data, columns=iris.feature_names)
y = iris.target
model = RandomForestClassifier().fit(X, y)
explainer = TabularExplainer(model, X, features=iris.feature_names)
global_explanation = explainer.explain_global()
print(global_explanation.feature_importance())