ml-assert
maintainedA lightweight, chainable assertion toolkit for validating data and models in ML workflows.
Python scikit-learn pandas pytest
ml-assert is a Python library that provides a fluent, expressive API to act as a guardrail in your automated ML pipelines.
The Problem
Standard metrics libraries just calculate scores. When an accuracy drops to 75%, the pipeline continues, propagating bad models.
The Solution
ml-assert asserts that accuracy must be >= 80%. If it’s 75%, it raises an AssertionError, halting execution. This is a paradigm shift from passive calculation to active assertion for robust ML Ops.
Features
- DataFrame Assertions
- Statistical Drift Detection
- Model Performance Assertions
- Extensible Plugin System
- Declarative CLI