Right now, the default argument is NULL, which relies on calculating every possible threshold. With large datasets and complex models that produce a high number of unique prediction scores, this can result in a large data frame being produced that causes a memory error.
We will instead change the defaults to something more reasonable involving an evenly spaced grid from 0 to 1.
Right now, the default argument is
NULL, which relies on calculating every possible threshold. With large datasets and complex models that produce a high number of unique prediction scores, this can result in a large data frame being produced that causes a memory error.We will instead change the defaults to something more reasonable involving an evenly spaced grid from 0 to 1.