Skip to contents

Inspect an Optimization Result

Source: R/optimization-result.R
optimization_result.Rd

optimization_result() is the stable, read-only boundary for learning what an optimizer did. Every dsprrr teleprompter reports the same core fields; optimizer-specific evidence lives under a namespaced extensions entry.

Usage

optimization_result(program)

# S3 method for class 'dsprrr_optimization_result'
print(x, ...)

Arguments

program

A dsprrr module returned by compile() or optimize_grid().

x

An optimization result.

...

Additional arguments, currently unused.

Value

A dsprrr_optimization_result with fields:

  • version: Result schema version.

  • optimizer: Optimizer identity.

  • status: Either "completed" or "partial".

  • baseline_score, best_score, and best_trial: Comparable outcome measures when the optimizer evaluates candidates.

  • best_params: Winning parameter values.

  • trials: Trial-level evidence as a tibble.

  • lineage: How the winning candidate was derived.

  • budget: Planned and consumed optimization budget.

  • stop_reason: Why the optimizer stopped.

  • extensions: Optimizer-specific evidence, namespaced by optimizer.

Returns NULL when program has not been optimized.

See also

Other optimizer accessors: apply_best_config(), best_demos(), best_params(), config_diff(), export_module_code(), optimization_summary(), top_trials()

Examples

if (FALSE) {
optimized <- compile(program, GEPA(metric = metric), trainset)
result <- optimization_result(optimized)
result$best_score
result$trials
result$extensions$gepa
}