opal model regression baseline — opal

A saved reference run for the WCPO bigeye tuna assessment that documents the objective, gradient, and RTMB report output used by the vignettes/baseline.Rmd workflow to detect numerical regressions and timing changes before and after refactors.

Usage

opal_baseline

Format

A named list accessible via data(opal_baseline) and packaged as data/opal_baseline.rda.

Source

vignettes/baseline.Rmd (the "Save or compare" chunk) which reproduces the reference run saved to data/opal_baseline.rda.

Details

Baseline contents

timestamp: Time when the baseline was created.
description: Free-form label ("Pre-refactoring baseline" by default).
opal_version: Version string of the installed opal package.
dimensions: Named list with n_year, n_age, n_fishery, n_season, n_len, n_lf, n_wt, and n_wf from the model inputs.
nll: Negative log-likelihood at the reference parameter set.
gradient: Gradient vector evaluated at obj$par for the baseline.
max_gr: Maximum absolute gradient entry.
timing: List with fn_elapsed, gr_elapsed, fn_median, and gr_median values that summarize runtime measurements used in the comparison vignette.
report: Sublist containing the baseline's number_ysa, spawning_biomass_y, catch and harvest rate predictions (catch_pred_fya, catch_pred_ysf, hrate_ysa, hrate_ysfa), selectivity curves (sel_fya), and log-likelihood components (lp_prior, lp_penalty, lp_rec, lp_cpue, lp_lf, lp_wf).
par_values: Vector of obj$par values that produced the baseline.

Examples

if (FALSE) { # \dontrun{
  data(opal_baseline)
  str(opal_baseline$dimensions)
  opal_baseline$timing$fn_median
} # }