フリー問題

問題文

A machine learning inference function processes rows one at a time and the team wants better throughput by handling many rows per call. Which change does the documentation describe for that?

選択肢

1. Register the function as a table function so that several rows can be emitted per call, which raises the throughput.
2. Have the handler receive batches of input rows as a pandas structure and return a result of the same length.
3. Have the handler open a second session inside itself so that several rows can be processed in parallel from within one call.
4. Increase the warehouse size and the cluster count for a bigger warehouse instead of changing the handler.