You want to reduce the total computation time required for a large SageMaker hyperparameter tuning job (previous smaller jobs took weeks). Which two actions will most reduce overall tuning computation time?

The Hyperband early-stopping strategy terminates poorly performing hyperparameter configurations early, preventing them from consuming unnecessary computational resources and significantly reducing overall tuning time. Setting a lower value for MaxNumberOfTrainingJobs directly limits the total number of training jobs run during the tuning process. Fewer jobs mean less computation, thus reducing the total time. Increasing the number of hyperparameters being tuned would likely increase, not decrease, computation time due to a larger search space. An exhaustive grid search is computationally expensive and would increase tuning time, especially for large search spaces. Setting a lower value for MaxParallelTrainingJobs would decrease parallelism, potentially increasing the total wall-clock time for the tuning job, as jobs would run sequentially rather than concurrently.