When resuming a sweep with Bayesian optimization, are the previous runs kept into consideration?

When I resume a sweep with method set to bayes (in its configuration), i.e. with wandb sweep --resume, does the Bayesian optimization process keep into consideration the parameter values already explored before the sweep was interrupted? Or is the previous history of the sweep ignored?

Hi @fantauzzi , whenever you pause a sweep and then resume it, all methods of a sweep retain history of the parameter values already explored. Resuming will pick up where the sweep left off and run to completion.

This topic was automatically closed 60 days after the last reply. New replies are no longer allowed.