Checkpointed Tuning Over Resamples (`tune_grid_checkpointed`)

A fingerprinted, resumable wrapper around [tune::tune_grid()] that evaluates tuning on a per-resample unit of work. Completed resample results are saved into `checkpoint_dir` and skipped on rerun. Returns an ordinary `tune_results` object so downstream functions like [tune::select_best()], [tune::collect_metrics()], [tune::finalize_workflow()], and [ggplot2::autoplot()] work identically.

Usage

tune_grid_checkpointed(
  object,
  resamples,
  grid = 10,
  metrics = NULL,
  control = tune::control_grid(),
  checkpoint_dir = NULL,
  shard = NULL,
  checkpoint_mismatch = c("recompute", "overwrite", "error"),
  progress = FALSE,
  ...
)

Arguments

object

A `workflow` or model specification.

resamples

An `rset` resampling object (e.g. from [rsample::vfold_cv()]).

grid

Number of grid points or a parameter grid data frame.

metrics

A `yardstick::metric_set()`.

control

A `tune::control_grid()` object.

checkpoint_dir

Directory to store per-resample fingerprinted checkpoints.

shard

Optional integer vector `c(i, n)` for distributed evaluation.

checkpoint_mismatch

Action when a checkpoint fingerprint mismatches: `"recompute"` (default), `"overwrite"`, or `"error"`.

progress

Logical (default `FALSE`).

...

Additional arguments passed to [tune::tune_grid()].

Value

A standard `tune_results` object.

See also

[fit_resamples_checkpointed()], [workflow_map_checkpointed()], [checkpoint_map()]