Core Concepts

These pages explain how DeepTab works under the hood: the scikit-learn interface, the split-config system, the training and evaluation pipeline, observability, and deployment-safe inference.

• sklearn API: The fit/predict/evaluate interface

• Model Tiers: Stable versus experimental models

• Custom Models: Building your own architectures

• Config System: Split configuration for model, preprocessing, and training

• Observability: Lifecycle events, structured logging, and experiment tracking

• Training and Evaluation: The fit pipeline, metrics, and reproducibility

• Model Operations: Serialisation and inspection

• Inference: Deployment-safe prediction with InferenceModel