Deepchecks
A Python toolkit for testing ML data and models, with reusable checks and production monitoring available through Docker or a hosted service.
Deepchecks is a Python toolkit for data scientists and ML teams checking datasets and models before deployment and watching for problems in production. Its testing, CI and monitoring components reuse the same checks across those stages.
- Data validation: Examine data integrity, distributions and train/test splits for tabular and text datasets, using built-in checks or your own test suites.
- Model evaluation: Compare models and assess performance to help decide whether a model is ready for production.
- Results and collaboration: Review visual results in Jupyter or HTML reports, process Python or JSON outputs in automated workflows, and use a shared interface to discuss test results before deployment.
- Production monitoring: Track deployed model behavior, detect data drift, receive alerts and investigate the causes of problems.
Monitoring can run on your own Linux or macOS server through Docker, or through Deepchecks' hosted SaaS service. The core testing and monitoring projects use AGPL 3.0. Some monitoring components carry a commercial license and require a Deepchecks license when enabled in a self-hosted instance. The separate Monitoring OSS repository excludes those proprietary components.
The 0.19.1 release distributes the Python validation library. Its open-source core uses AGPL v3; enterprise components have separate commercial terms. This listing covers the local library rather than the hosted commercial product.
