試験
通常
試験時間: 00:00問題時間: 00:00
問題一覧
問題 32 / 65
Which strategy is optimal for reducing inference costs while maintaining accuracy in a multilingual translation service with a large number of simultaneous connections?
①Extremely quantize the model to make it lightweight regardless of accuracy.
②Process all requests sequentially using the same large model.
③Do not perform any caching or batch processing at all.
④Run a separate inference with the full-size model for each user every time.
⑤Create lightweight models through model distillation and operate in a hybrid mode that routes to larger models as needed.