AMake a trained model available to answer live requests in production
BTrain a model on fresh data until its accuracy stops improving more
CCompress a model so its stored weights take far less space on disk
DEvaluate a model on a held-out dataset to measure its final quality
Why this is the answer
Serving exposes an already-trained model so clients can send inputs and get outputs — it is the inference stage, distinct from training, compression, or evaluation.