int_conformal_cv doesn't work with group_vfold_cv

[Jeffrothschild]

The problem

I'm having trouble using int_conformal_cv when the resamples were created from group_vfold_cv()

Here is an example I modified from https://www.tidymodels.org/learn/models/conformal-regression/#using-resampling-results

Reproducible example

make_data <- function(n, std_dev = 1 / 5) {
  tibble(x = runif(n, min = -1)) %>%
    mutate(
      y = (x^3) + 2 * exp(-6 * (x - 0.3)^2),
      y = y + rnorm(n, sd = std_dev)
    )
}

n <- 1000
set.seed(8383)
train_data <- make_data(n) %>% 
  mutate(color = sample(c('red', 'blue'), n(), replace = TRUE))

set.seed(7292)
test_data <- make_data(10000) %>% 
  mutate(color = sample(c('red', 'blue'), n(), replace = TRUE))

set.seed(493)
folds <- vfold_cv(train_data)
group_folds <- group_vfold_cv(train_data, group = color)

set.seed(484)
nnet_wflow <- 
  workflow(y ~ x, mlp(hidden_units = 4) %>% set_mode("regression"))

ctrl <- control_resamples(save_pred = TRUE, extract = I)

nnet_rs <- 
  nnet_wflow %>% 
  fit_resamples(folds, control = ctrl)

collect_metrics(nnet_rs)

# This works

cv_int <- int_conformal_cv(nnet_rs)

predict(cv_int, test_data, level = 0.90) %>% bind_cols(test_data)

# try again with group cv and get an error

group_nnet_rs <- 
  nnet_wflow %>% 
  fit_resamples(group_folds, control = ctrl)

collect_metrics(group_nnet_rs)

group_cv_int <- int_conformal_cv(group_nnet_rs)

# Error in if (rs$att$class != "vfold_cv") { : the condition has length > 1


reprex::reprex(si = TRUE)

[topepo]

There's a PR to fix it. Thanks for reporting this.

It will still issue a warning though.