In xgboost, `ntreelimit` has been deprecated in favour of `iteration_range`

[tiagomaie]

parsnip/R/boost_tree.R

Lines 484 to 486 in 0af9c86

	xgb_by_tree <- function(tree, object, new_data, type, ...) {
	pred <- xgb_pred(object$fit, newdata = new_data, ntreelimit = tree)

L485 should instead be:
pred <- xgb_pred(object$fit, newdata = new_data, iterationrange = c(1, tree + 1), ntreelimit = NULL)

ntreelimit has been deprecated.
It is tree + 1 as in the xgboost docs they mention that iteration_range should be a half-open set.
This should fix the warnings given when running tune::tune_grid for example, see below.

library(parsnip)
library(dplyr)

# using iris dataset

# defining train/test split
train_ids <- 1:(nrow(iris) - 50)
test_ids <- (nrow(iris) - 49):nrow(iris)

# defining recipe
d_recipe <- recipes::recipe(
  Species ~ .,
  data = iris[train_ids, ] %>% dplyr::mutate(Species = factor(Species))
)

# defining model
g_model <-
  parsnip::boost_tree(trees = tune::tune()) %>%
  parsnip::set_engine("xgboost", objective = "binary:logistic") %>%
  parsnip::set_mode("classification")


# defining workflow
wf <- workflows::workflow() %>%
  workflows::add_recipe(d_recipe) %>%
  workflows::add_model(g_model)

# tuning params
res <- wf %>% tune::tune_grid(
    resamples=rsample::vfold_cv(
      recipes::bake(d_recipe %>% recipes::prep(), new_data = NULL)
    )
  )
#> Registered S3 method overwritten by 'tune':
#>   method                   from   
#>   required_pkgs.model_spec parsnip
#> [18:38:53] WARNING: amalgamation/../src/c_api/c_api.cc:718: `ntree_limit` is deprecated, use `iteration_range` instead.
#> [18:38:53] WARNING: amalgamation/../src/c_api/c_api.cc:718: `ntree_limit` is deprecated, use `iteration_range` instead.
#> [18:38:53] WARNING: amalgamation/../src/c_api/c_api.cc:718: `ntree_limit` is deprecated, use `iteration_range` instead.
...
#> [18:38:59] WARNING: amalgamation/../src/c_api/c_api.cc:718: `ntree_limit` is deprecated, use `iteration_range` instead.
#> [18:38:59] WARNING: amalgamation/../src/c_api/c_api.cc:718: `ntree_limit` is deprecated, use `iteration_range` instead.
#> [18:38:59] WARNING: amalgamation/../src/c_api/c_api.cc:718: `ntree_limit` is deprecated, use `iteration_range` instead.

^{Created on 2022-02-09 by the reprex package (v2.0.1)}

I should note that this potential fix doesn't account for functionality that you can have with iterationrange but that it is currently not possible anyway with ntreelimit.

[juliasilge]

It looks like this was part of v1.5.0:

In 1.4 we have a new prediction function in the C API which is used by the Python package. This release revises the R package to use the new prediction function as well. A new parameter iteration_range for the predict function is available, which can be used for specifying the range of trees for running prediction. (#6819, #7126)

And I believe this was released on CRAN in November 2021.

[github-actions]

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.