After model training is completed, the platform provides several key insights and artifacts for the model, including:
- Final model accuracy
-
Accuracies and footprints of models at each iteration
Note
To download a smaller model and tolerate slightly lower accuracy, select any iteration from the chart. All metrics, footprints, and analytics will be recalculated automatically, and a corresponding archive will be prepared for download.
-
Model footprint for the selected target hardware. If multiple devices were selected, use the Target hardware dropdown to choose another device. The displayed footprint will update accordingly.
-
Data Analytics, that shows the distribution of variables in the dataset.
-
Model Quality Diagram, that represents multiple metrics relevant to the task type.
- Feature Importance Matrix (FIM), that shows the contribution of each input feature to the model's predictions.
-
Confusion Matrix, that visualizes the model's performance by showing correct and incorrect predictions for each class.
You can download the archive to your PC.