Extends the sibling HPO campaign's own ResumableTrial (successive_halving.hpp) with the weight/hyperparameter read-write capability Population Based Training's own exploit/explore step needs. More...

#include <vector>
#include "pulsatrix/search_space.hpp"
#include "pulsatrix/successive_halving.hpp"

Include dependency graph for pbt_trial.hpp:

This graph shows which files directly or indirectly include this file:

Go to the source code of this file.

Classes
class	pulsatrix::PBTResumableTrial
	A ResumableTrial that additionally exposes its live weights (a flat vector) and its current hyperparameter Configuration, both readable and replaceable mid-training. More...

Namespaces
namespace	pulsatrix

Detailed Description

Extends the sibling HPO campaign's own ResumableTrial (successive_halving.hpp) with the weight/hyperparameter read-write capability Population Based Training's own exploit/explore step needs.

Note: Recon finding (this mission's own activation check, 2026-09-27): ResumableTrial exists and is real (built for HPO Phase 3's Successive Halving/Hyperband/ASHA), but exposes only TrainForEpochs(int) -> double – exactly as successive_halving.hpp's own note predicted, it does not suffice for PBT's exploit step (copying a top performer's live weights and hyperparameters onto a bottom performer mid-training) as-is. Per this project's own precedent (RL's Agent interface extended per-consumer – DQNAgent's epsilon-greedy methods, CategoricalPolicyAgent's log-probability accessor – rather than speculatively upfront), this is a new, purely-additive derived interface. Zero changes to ResumableTrial itself: Successive Halving/Hyperband/ASHA are completely unaffected.

Classes

Namespaces

Detailed Description