transition parsnip::set_model_args() inputs to main arguments

[simonpcouch]

Hey there! Super stoked yall are putting this together.

Similar reprex to tidymodels/parsnip#832, with this PR:

library(tidybert)
library(tidymodels)
library(modeldata)
tidymodels_prefer()

data(tate_text)
tate_text <- tate_text %>%
  # We'll just work with the top 6 artists.
  dplyr::filter(
    artist %in% c(
      "Schütte, Thomas", "Zaatari, Akram", "Beuys, Joseph", "Ferrari, León",
      "Kossoff, Leon", "Tillmans, Wolfgang"
    )
  ) %>%
  dplyr::mutate(
    artist = as.factor(as.character(artist))
  )
set.seed(4242)
tate_folds <- rsample::vfold_cv(tate_text, v = 3)

tidybert_spec <-
  bert(
    mode = "classification",
    epochs = 1,
    bert_type = tune(),
    n_tokens = tune()
  ) %>%
  parsnip::set_engine(
    "tidybert"
  )

tidybert_workflow <- workflows::workflow(
  artist ~ title,
  tidybert_spec
)

grid <- tidybert_workflow %>%
  workflows::extract_parameter_set_dials() %>%
  update(
    bert_type = bert_type(
      c("bert_small_uncased", "bert_tiny_uncased")
    ),
    n_tokens = n_tokens(c(4, 5))
  ) %>%
  dials::grid_latin_hypercube(size = 2)

torch::torch_manual_seed(4242)
tidybert_result <- tidybert_workflow %>%
  tune::tune_grid(
    resamples = tate_folds,
    grid = grid
  )

tidybert_result
#> # Tuning results
#> # 3-fold cross-validation 
#> # A tibble: 3 × 4
#>   splits            id    .metrics         .notes          
#>   <list>            <chr> <list>           <list>          
#> 1 <split [405/203]> Fold1 <tibble [4 × 6]> <tibble [0 × 3]>
#> 2 <split [405/203]> Fold2 <tibble [4 × 6]> <tibble [0 × 3]>
#> 3 <split [406/202]> Fold3 <tibble [4 × 6]> <tibble [0 × 3]>

^{Created on 2022-10-28 with reprex v2.0.2}

Some more details on this suggestion here. :)

[CLAassistant]

All committers have signed the CLA.

[jonthegeek]

Thank you so much! I'm merging this into a branch where I'm adding checks and vignette details around this, rather than into main, but this will be a huge help!