DNIKit 2.0.0 Release

The Data and Network Introspection Kit (DNIKit) has officially been open sourced! DNIKit is a Python toolkit for analyzing machine learning models and datasets, and can be used to:

• create a comprehensive dataset analysis report
• identify duplicate data samples
• find rare data samples, annotation errors, or model biases
• compress networks
• detect inactive units in a model (e.g., due to dying ReLU)
• …and more.

To learn more and get started, check out the DNIKit Documentation.