experiment-compare

You can compare MLflow experiment runs by providing specific run IDs, run names, or selecting the latest N runs. The tool aggregates metrics, hyperparameters, and artifacts into a structured side-by-side report to justify the best model choice.

You can compare metrics such as accuracy, F1, AUC, and RMSE across MLflow experiment runs. The comparison also aggregates hyperparameters, feature sets, and training characteristics to identify what drives model performance.

Yes, you can include optional artifacts like feature importance files or confusion matrices when evaluating MLflow runs. These artifacts are aggregated alongside metrics and hyperparameters to provide a comprehensive model evaluation.

You need access to an MLflow tracking server with existing experiment runs that have logged metrics and parameters. Optional artifacts like feature importance files or confusion matrices can be included if available.

The best way to document MLflow model selection decisions is to generate a reproducible comparison notebook alongside a concise recommendation. This aggregates run metrics and hyperparameters into a structured report for stakeholders.