問題文
A platform team is planning the capacity of a new AI environment and wants to understand where the AI lifecycle actually consumes infrastructure. Which statement reflects the lifecycle most accurately?
選択肢
- The lifecycle begins with model training, because there is nothing to plan for until the accelerators are allocated and the first run starts.
- Every stage of the lifecycle places the same demand on the platform, so a single uniform pool of accelerators sized for training is sufficient for the whole cycle.
- Data collection and preparation usually consume a large share of the elapsed time and of the storage capacity, well before any accelerator is busy.
- The lifecycle ends when the model is deployed, so monitoring and retraining are activities that belong to a different discipline and do not consume platform capacity.