Workflow size warning threshold

NEWS.md

@@ -1,5 +1,7 @@
 # tune (development version)
 
+* The warning threshold when check the size of a workflow is now a parameter to the control functions and has a new default of 100MB. (#914)
+
 # tune 2.0.1
 
 * Fixed a bug where `int_pctl()` wouldn't work on `last_fit()` outcomes when future parallelism was enabled. (#1099) 

R/control.R

@@ -37,7 +37,8 @@ control_grid <- function(
   save_workflow = FALSE,
   event_level = "first",
   parallel_over = NULL,
-  backend_options = NULL
+  backend_options = NULL,
+  workflow_size = 100.0
 ) {
   # Any added arguments should also be added in superset control functions
   # in other packages
@@ -50,6 +51,7 @@ control_grid <- function(
   check_string(event_level)
   check_character(pkgs, allow_null = TRUE)
   check_function(extract, allow_null = TRUE)
+  check_number_decimal(workflow_size)
 
   val_parallel_over(parallel_over, "control_grid()")
 
@@ -62,7 +64,8 @@ control_grid <- function(
     save_workflow = save_workflow,
     event_level = event_level,
     parallel_over = parallel_over,
-    backend_options = backend_options
+    backend_options = backend_options,
+    workflow_size = workflow_size
   )
 
   class(res) <- c("control_grid", "control_resamples")
@@ -200,6 +203,9 @@ print.control_last_fit <- function(x, ...) {
 #'   backend. Defaults to `NULL` for default backend options.
 #' @param allow_par A logical to allow parallel processing (if a parallel
 #'   backend is registered).
+#' @param workflow_size A non-negative number that is used as a threshold for a
+#' warning regarding the size of the workflow. Only used when
+#' `save_workflow = TRUE`.
 #'
 #' @inheritSection collect_predictions Hyperparameters and extracted objects
 #'
@@ -240,7 +246,8 @@ control_bayes <-
     event_level = "first",
     parallel_over = NULL,
     backend_options = NULL,
-    allow_par = TRUE
+    allow_par = TRUE,
+    workflow_size = 100.0
   ) {
     # Any added arguments should also be added in superset control functions
     # in other packages
@@ -257,6 +264,7 @@ control_bayes <-
     check_number_whole(no_improve, min = 0, allow_infinite = TRUE)
     check_number_whole(uncertain, min = 0, allow_infinite = TRUE)
     check_number_whole(seed)
+    check_number_decimal(workflow_size)
 
     check_time_limit_arg(time_limit)
 
@@ -285,7 +293,8 @@ control_bayes <-
         save_gp_scoring = save_gp_scoring,
         event_level = event_level,
         parallel_over = parallel_over,
-        backend_options = backend_options
+        backend_options = backend_options,
+        workflow_size = workflow_size
       )
 
     class(res) <- "control_bayes"

R/tune_grid.R

@@ -439,7 +439,7 @@ set_workflow <- function(workflow, control) {
     if (!is.null(workflow$pre$actions$recipe)) {
       w_size <- utils::object.size(workflow$pre$actions$recipe)
       # make 5MB cutoff
-      if (w_size / 1024^2 > 5) {
+      if (w_size / 1024^2 > control$workflow_size) {
         msg <-
           paste0(
             "The workflow being saved contains a recipe, which is ",

tests/testthat/_snaps/control.md

@@ -0,0 +1,11 @@
+# workflow size warning
+
+    Code
+      set.seed(1)
+      warns <- fit_resamples(lm_wflow, resamples = vfold_cv(MTCARS), control = control_resamples(
+        save_workflow = TRUE, workflow_size = 2))
+    Message
+      i The workflow being saved contains a recipe, which is 2.7 Mb in i memory. If
+      this was not intentional, please set the control setting i `save_workflow =
+      FALSE`.
+

tests/testthat/test-control.R

@@ -0,0 +1,38 @@
+test_that("workflow size warning", {
+  # A larger data set to trip the warning
+  MTCARS <- mtcars[rep(1:32, each = 1000), ]
+
+  lm_rec <- recipe(mpg ~ ., data = MTCARS)
+  lm_wflow <- workflow(lm_rec, linear_reg() |> set_engine("lm", x = TRUE))
+  # About 2.7MB when fit
+
+  expect_silent({
+    set.seed(1)
+    no_warning <-
+      lm_wflow |>
+      fit_resamples(
+        resamples = vfold_cv(MTCARS),
+        control = control_resamples(save_workflow = TRUE, workflow_size = Inf)
+      )
+  })
+
+  expect_snapshot({
+    set.seed(1)
+    warns <-
+      lm_wflow |>
+      fit_resamples(
+        resamples = vfold_cv(MTCARS),
+        control = control_resamples(save_workflow = TRUE, workflow_size = 2)
+      )
+  })
+
+  expect_silent({
+    set.seed(1)
+    no_save <-
+      lm_wflow |>
+      fit_resamples(
+        resamples = vfold_cv(MTCARS),
+        control = control_resamples(save_workflow = FALSE, workflow_size = 2)
+      )
+  })
+})

tests/testthat/test-grid.R

@@ -786,20 +786,4 @@ test_that("retain extra attributes", {
   )
   expect_null(attr(res, "workflow"))
   expect_true(inherits(attr(res2, "workflow"), "workflow"))
-
-  wflow2 <- workflow() |>
-    add_recipe(recipes::recipe(mpg ~ ., mtcars[rep(1:32, 3000), ])) |>
-    add_model(helper_objects$svm_mod)
-  pset2 <- extract_parameter_set_dials(wflow2)
-  grid2 <- dials::grid_regular(pset2, levels = 3)
-
-  expect_message(
-    tune_grid(
-      wflow2,
-      resamples = folds,
-      grid = grid2,
-      control = control_grid(save_workflow = TRUE)
-    ),
-    "being saved contains a recipe, which is"
-  )
 })

tests/testthat/test-resample.R

@@ -484,15 +484,6 @@ test_that("retain extra attributes", {
     )
   expect_null(attr(res, "workflow"))
   expect_true(inherits(attr(res2, "workflow"), "workflow"))
-
-  expect_snapshot(
-    fit_resamples(
-      lin_mod,
-      recipes::recipe(mpg ~ ., mtcars[rep(1:32, 3000), ]),
-      folds,
-      control = control_resamples(save_workflow = TRUE)
-    )
-  )
 })