Emil Hvitfeldt

We need to isolate the how options are being set. the following test https://github.com/tidymodels/tune/blob/05242209934d651de6198e5c488051f1dc45a5c4/tests/testthat/test-bayes.R#L419-L466 is being affect by the option being set in the previous test https://github.com/tidymodels/tune/blob/05242209934d651de6198e5c488051f1dc45a5c4/tests/testthat/test-bayes.R#L374

We are catching this error differently now using the `no_penalty()` function. The error pops up in `submod_and_others()` if the `no_penalty()` check is ignored. ``` r library(tidymodels) rec_no_tune_1 % step_normalize(all_predictors()) glmn_mod...

Hello @Tripartio 👋 welcome! I think the issue you are running into, is happening because of a typo. When using `roc_auc_survival()` you are using `.pred` but then when you try...

> Here is the documentation for the estimate argument for roc_auc_survival_vec(): Thank you, that is a mistake, and should be fixed. As a general rule in {yardstick}, all metrics that...

Also, i hope this doesn't come off condersending, as it isn't the point 🤞 The way you calculate the ROC curve for censored data is different than how you calculate...

Hello @EduMinsky 👋 You are getting these results because by default {yardstick} uses the first factor level (`N_Activated` in this case) as the basis for calculating metrics. This means that...

Removing `extras` makes sense to me. according to `?axe_data.workflow` it reads: > After calling butcher::butcher() on a workflow, the only guarantee is that you will still be able to predict()...

@simonpcouch should be able to confirm. But I'm pretty sure we don't need them for prediction

This seems like a reasonable thing to do. Should properly live in the `*_metric_summarizer()` functions as a `check_na_rm()` call

related to https://github.com/tidymodels/yardstick/issues/419