Model Evaluation

Basic Evaluation

uv run python -m mlops_project.evaluate --checkpoint models/best_model.ckpt

Evaluation Metrics

The evaluation produces: - Accuracy: Overall classification accuracy - Precision/Recall: Per-class metrics - F1 Score: Harmonic mean - Confusion Matrix: Prediction distribution - AUC-ROC: Classification threshold analysis

Output

Results are saved to outputs/evaluation/: - metrics.json - Numerical results - confusion_matrix.png - Visualization - roc_curve.png - ROC analysis