フリー問題

問題文

A practitioner replaces a scikit-learn linear regression with the cuML equivalent. What does the documentation say about the shape of the resulting code?

選択肢

1. The estimator must be compiled into a device-specific plan before it can be fitted, and the plan has to be rebuilt whenever the number of rows changes, which is why the interface adds a preparation step that has no counterpart on the processor side.
2. Inference requires re-fitting the estimator on the test split.
3. The estimator is constructed with parameters, trained with a fitting call, and used for inference with a prediction call, mirroring the familiar interface.
4. The data must be converted to a plain list before fitting.