caret binary classification eats up the CPU #1413
Description
I am trying to share a high performance computer with a group of students. I have one generic script that is written to cover regression, binary-classification, and multiclass-classification. I have explictly set allowParallel=FALSE. summaryFunction = mnLogLoss for both classification.
What I find is that binary classification eats up the entire CPU. This is ironic as I would have expected all three to fight for the same resource equally. Additionally, if I am explicitly setting allowParallel=FALSE, confused why caret still uses all cores.
How I can explicitly force caret to use 1-2 cores so that multiple parallel scripts share the same computer from a single TA's login account?
Prefer R solution instead a unix solution.
Hope someone can guide.