ML Diagnostics are designed to identify and help troubleshoot potential problems and suggest possible improvements at different stages of training and building machine learning models.
Assertions are not available for clustering
Assertions are not available for Deep learning models
Assertions are not available for computer vision models
Assertions are not available for time series forecasting models
Assertions are not available for causal prediction models
An ML assertion will help validate that the machine learning model outputs the expected class for given examples.
Define those examples and rules then check whether models respect them after training. This does not aim at enforcing model predictions depending on some rules. It is a validation tool.
Valid ratio will not consider rows dropped by the model.