フリー問題

問題文

What is the primary reason for placing inference capacity at the edge rather than in a central data center?

選択肢

1. To centralize monitoring, since edge locations report a single combined health value and therefore need less operational attention than a central site.
2. To make training faster, because the edge nodes can be joined into the central training cluster and contribute their accelerators to the same collective operations.
3. To reduce the accelerator memory each model needs, since edge models are compressed, so the same model would fit a smaller accelerator after the move.
4. To keep the round trip short and to keep working even when the link back to the central site is slow or unavailable.