LightGBM grows leaf-wise, so it's possible to get a very unbalanced tree. Depending on your data, it's completely possible to end up with a tree that has fewer than 2^17 leaves but a depth much greater than 17.
I think the lines below are intended to cap the number of leaves at the max (2^17), but they're actually capping the depth, NOT the total number of leaves. This seems like a fairly critical bug since you're inadvertently limiting the possible depth of trees.
LightGBM grows leaf-wise, so it's possible to get a very unbalanced tree. Depending on your data, it's completely possible to end up with a tree that has fewer than 2^17 leaves but a depth much greater than 17.
I think the lines below are intended to cap the number of leaves at the max (2^17), but they're actually capping the depth, NOT the total number of leaves. This seems like a fairly critical bug since you're inadvertently limiting the possible depth of trees.
https://github.com/curso-r/treesnip/blob/3e4c2207cd137f1764b10ea019cea2218875bc48/R/lightgbm.R#L258-L261
Edit: Just realized that
max_depthis never assigned back toarg_listso it doesn't actually cap anything it, just throws a warning.