Evidently

An open-source framework to evaluate, test and monitor ML models in production.

Docs | Discord Community | Newsletter | Blog | Twitter

📊 What is Evidently?

Evidently helps analyze and track data and ML model quality throughout the model lifecycle. You can think of it as an evaluation layer that fits into the existing ML stack.

Evidently has a modular approach with 3 interfaces on top of the shared analyzer functionality.

1. Interactive visual reports

Evidently generates interactive dashboards from pandas DataFrame or csv files. You can use them for model evaluation, debugging and documentation.

Each report covers a particular aspect of the model performance. You can display reports in Jupyter notebook or Colab or export as an HTML file. Currently 7 pre-built reports are available:

Data Drift. Detects changes in the input feature distribution.
Data Quality. Provides the detailed feature stats and behavior overview.
Target Drift: Numerical, Categorical. Detects changes in the model output.
Model Performance: Classification, Probabilistic Classification, Regression. Evaluates the quality of the model and model errors.

2. Data and ML model profiling

Evidently also generates JSON profiles. You can use them to integrate the data or model evaluation step into the ML pipeline.

You can log and store JSON profiles for further analysis, or build a conditional workflow based on the result of the check (e.g. to trigger alert, retraining, or generate a visual report). The profiles calculate the same metrics and statistical tests as visual reports.

You can explore example integrations with tools like Airflow and Mlflow.

3. Real-time ML monitoring

Note: this functionality is in active development and subject to API change.

Evidently has monitors that collect the data and model metrics from a deployed ML service. You can use it to build live monitoring dashboards. Evidently configures the monitoring on top of the streaming data and emits the metrics.

There is a lightweight integration with Prometheus and Grafana that comes with pre-built dashboards.

👩‍💻 Installing from PyPI

MAC OS and Linux

Evidently is available as a PyPI package. To install it using pip package manager, run:

$ pip install evidently

If you want to generate reports as HTML files or export as JSON profiles, the installation is now complete.

If you want to display the dashboards directly in a Jupyter notebook, you should install jupyter nbextension. After installing evidently, run the two following commands in the terminal from the evidently directory.

To install jupyter nbextension, run:

$ jupyter nbextension install --sys-prefix --symlink --overwrite --py evidently

To enable it, run:

$ jupyter nbextension enable evidently --py --sys-prefix

That's it! A single run after the installation is enough.

Note: if you use Jupyter Lab, the dashboard might not display in the notebook. However, the report generation in a separate HTML file will work correctly.

Windows

Evidently is available as a PyPI package. To install it using pip package manager, run:

$ pip install evidently

The tool allows building interactive reports both inside a Jupyter notebook and as a separate HTML file. Unfortunately, building reports inside a Jupyter notebook is not yet possible for Windows. The reason is Windows requires administrator privileges to create symlink. In later versions we will address this issue.

▶️ Getting started

Jupyter Notebook

To start, prepare your data as two pandas DataFrames. The first should include your reference data, the second - current production data. The structure of both datasets should be identical.

For Data Drift report, include the input features only.
For Target Drift reports, include the column with Target and/or Prediction.
For Model Performance reports, include the columns with Target and Prediction.

Calculation results can be available in one of the two formats:

Option 1: an interactive Dashboard displayed inside the Jupyter notebook or exportable as a HTML report.
Option 2: a JSON Profile that includes the values of metrics and the results of statistical tests.

Option 1: Dashboard

After installing the tool, import Evidently dashboard and required tabs:

import pandas as pd
from sklearn import datasets

from evidently.dashboard import Dashboard
from evidently.dashboard.tabs import (
    DataDriftTab,
    CatTargetDriftTab
)

iris = datasets.load_iris()
iris_frame = pd.DataFrame(iris.data, columns = iris.feature_names)
iris_frame['target'] = iris.target

To generate the Data Drift report, run:

iris_data_drift_report = Dashboard(tabs=[DataDriftTab()])
iris_data_drift_report.calculate(iris_frame[:100], iris_frame[100:], column_mapping = None)
iris_data_drift_report.save("reports/my_report.html")

To generate the Data Drift and the Categorical Target Drift reports, run:

iris_data_and_target_drift_report = Dashboard(tabs=[DataDriftTab(), CatTargetDriftTab()])
iris_data_and_target_drift_report.calculate(iris_frame[:100], iris_frame[100:], column_mapping = None)
iris_data_and_target_drift_report.save("reports/my_report_with_2_tabs.html")

If you get a security alert, press "trust html". HTML report does not open automatically. To explore it, you should open it from the destination folder.

Option 2: Profile

After installing the tool, import Evidently profile and required sections:

import pandas as pd
from sklearn import datasets

from evidently.model_profile import Profile
from evidently.model_profile.sections import (
    DataDriftProfileSection,
    CatTargetDriftProfileSection
)

iris = datasets.load_iris()
iris_frame = pd.DataFrame(iris.data, columns = iris.feature_names)

To generate the Data Drift profile, run:

iris_data_drift_profile = Profile(sections=[DataDriftProfileSection()])
iris_data_drift_profile.calculate(iris_frame, iris_frame, column_mapping = None)
iris_data_drift_profile.json()

To generate the Data Drift and the Categorical Target Drift profile, run:

iris_target_and_data_drift_profile = Profile(sections=[DataDriftProfileSection(), CatTargetDriftProfileSection()])
iris_target_and_data_drift_profile.calculate(iris_frame[:75], iris_frame[75:], column_mapping = None) 
iris_target_and_data_drift_profile.json()

Google Colab, Kaggle Kernel, Deepnote

Read instructions on how to run Evidently in other notebook environments.

You can run evidently in Google Colab, Kaggle Notebook and Deepnote.

First, install evidently. Run the following command in the notebook cell: !pip install evidently

There is no need to enable nbextension for this case, because evidently uses an alternative way to display visuals in the hosted notebooks.

To build a Dashboard or a Profile simply repeat the steps described in the previous paragraph. For example, to build the Data Drift dashboard, run:

import pandas as pd
from sklearn import datasets

from evidently.dashboard import Dashboard
from evidently.dashboard.tabs import DataDriftTab

iris = datasets.load_iris()
iris_frame = pd.DataFrame(iris.data, columns = iris.feature_names)

iris_data_drift_report = Dashboard(tabs=[DataDriftTab()])
iris_data_drift_report.calculate(iris_frame[:100], iris_frame[100:], column_mapping = None)

To display the dashboard in the Google Colab, Kaggle Kernel, Deepnote, run:

iris_data_drift_report.show()

The show() method has the argument mode, which can take the following options:

auto - the default option. Ideally, you will not need to specify the value for mode and use the default. But, if it does not work (in case we failed to determine the environment automatically), consider setting the correct value explicitly.
nbextension - to show the UI using nbextension. Use this option to display dashboards in Jupyter notebooks (it should work automatically).
inline - to insert the UI directly into the cell. Use this option for PyLab, Google Colab, Kaggle Kernels and Deepnote. For Google Colab, this should work automatically, for PyLab, Kaggle Kernels and Deepnote the option should be specified explicitly.

💻 Contributions

We welcome contributions! Read the Guide to learn more.

🖼️ Community Reports

You can also contribute custom reports with a combination of own metrics and widgets. We'll be glad to showcase some of them!

A simple dashboard which contains two custom widgets with target distribution information: link to repository

📚 Documentation

For more information, refer to a complete Documentation. You can start with this Tutorial for a quick introduction.

🗂️ Examples

Here you can find simple examples on toy datasets to quickly explore what Evidently can do right out of the box.

Report	Jupyter notebook	Colab notebook	Data source
Data Drift + Categorical Target Drift (Multiclass)	link	link	Iris plants sklearn.datasets
Data Drift + Categorical Target Drift (Binary)	link	link	Breast cancer sklearn.datasets
Data Drift + Numerical Target Drift	link	link	California housing sklearn.datasets
Regression Performance	link	link	Bike sharing UCI: link
Classification Performance (Multiclass)	link	link	Iris plants sklearn.datasets
Probabilistic Classification Performance (Multiclass)	link	link	Iris plants sklearn.datasets
Classification Performance (Binary)	link	link	Breast cancer sklearn.datasets
Probabilistic Classification Performance (Binary)	link	link	Breast cancer sklearn.datasets
Data Quality	link	link	Bike sharing UCI: link

Integrations

See how to integrate Evidently in your prediction pipelines and use it with other tools.

Title	link to tutorial
Real-time ML monitoring with Grafana	Evidently + Grafana
Batch ML monitoring with Airflow	Evidently + Airflow
Log Evidently metrics in MLflow UI	Evidently + MLflow

☎️ Community Call

We host monthly community call for users and contributors. Sign up to join the next one.

✅ Stay updated

If you want to receive updates, follow us on Twitter, or sign up for our newsletter.
You can also find more tutorials and explanations in our Blog.
If you want to chat and connect, join our Discord community!

jeff-ridgeway/evidently