pass along nfeatures and feature_names to xgboost bundle

[EmilHvitfeldt]

Some functions that are applied to xgboost models use the feature_names and/or nfeatures element of the model. This PR makes sure they aren't removed when bundling.

Should close #66 and deal with tidymodels/orbital#61

library(tidymodels)
library(bundle)
library(orbital)

mod <- boost_tree(trees = 5, mtry = 3) %>%
  set_mode("regression") %>%
  set_engine("xgboost") %>%
  fit(mpg ~ ., data = mtcars[1:25,])

mod$fit$feature_names
#>  [1] "cyl"  "disp" "hp"   "drat" "wt"   "qsec" "vs"   "am"   "gear" "carb"

vip::vi(mod)
#> # A tibble: 4 × 2
#>   Variable Importance
#>   <chr>         <dbl>
#> 1 hp           0.498 
#> 2 wt           0.213 
#> 3 disp         0.212 
#> 4 drat         0.0774

mod_bundle <- bundle(mod)
mod_new <- unbundle(mod_bundle)

mod_new$fit$feature_names
#>  [1] "cyl"  "disp" "hp"   "drat" "wt"   "qsec" "vs"   "am"   "gear" "carb"

vip::vi(mod_new)
#> # A tibble: 4 × 2
#>   Variable Importance
#>   <chr>         <dbl>
#> 1 hp           0.498 
#> 2 wt           0.213 
#> 3 disp         0.212 
#> 4 drat         0.0774

^{Created on 2024-11-07 with reprex v2.1.0}

[juliasilge]

Would you want to update the tests as well, to confirm these components are usable after bundling?

[EmilHvitfeldt]

Done!

[juliasilge]

Thank you so much!