All models failed in tune_grid() about tune HOT 6 CLOSED

I think I have found the cause of this - I did not have kernlab installed.

With a bit of debug()-ing, I got to rs <- rlang::eval_tidy(code_path) inside tune_grid_workflow().

There, code_path evaluated to tune_mod_with_formula(rs, grid, object, perf, control), and debug() of tune_mod_with_formula() showed me errors due to kernlab not being found.

Installing kernlab has resolved the issue, and I can now reproduce the output in the article/vignette. This may be related to #44, although I'm very sorry that I'm not currently able to offer a solution to that issue, though.

from tune.

It's just been pointed out to me that kernlab is in Suggests (as in #59), so maybe it needs to be moved to Imports if it's needed for tuning? (Or maybe I just need to be smarter about how I install stuff...)

from tune.

It's only needed for tuning models that need it, though. I can understand not making someone install xgboost, for example, if they don't actually need xgboost.

I think better error messages would solve this, as in #44. Just need to figure out how to get there :)

from tune.

Yeah that's fair. Pretty sure at least part of this issue is due to a pebkac.

@topepo you asked in #59 if we were using parallel processing. I wasn't explicitly asking for it, but as I was passing ctrl <- grid_control(verbose = FALSE) to tune_grid() and the default for allow_par in grid_control() is TRUE, there's a chance that I was.

Either way, I think we've identified what was going wrong here, and it looks like #63 will resolve it.

from tune.

And having read the parallel section of the docs https://tidymodels.github.io/tune/articles/extras/optimizations.html#parallel-processing a bit more closely, I think the answer is: no, I'm not using parallel processing.

from tune.

This issue has been automatically locked. If you believe you have found a related problem, please file a new issue (with a reprex: https://reprex.tidyverse.org) and link to this issue.

from tune.