def ask(self, idg: solver.TrialIdGenerator) -> solver.NextTrial:
if not self._pruned.empty():
kurobako_trial_id, trial = self._pruned.get()
next_step = None
elif self._waitings.empty():
kurobako_trial_id = idg.generate()
trial = self._study.ask()
next_step = self._next_step(0)
else:
kurobako_trial_id, trial = self._waitings.get()
# TODO(contramundum53): remove access to self._study._storage
current_step = self._study._storage.get_trial(
trial._trial_id).last_step
next_step = self._next_step(current_step)
params = [] # type: List[Optional[float]]
for p in self._problem.params:
if p.is_constraint_satisfied(self._problem.params, params):
params.append(self._suggest(trial, p))
else:
params.append(None)
self._runnings[kurobako_trial_id] = trial
return solver.NextTrial(trial_id=kurobako_trial_id,
params=params,
next_step=next_step)
def ask(self, idg: solver.TrialIdGenerator) -> solver.NextTrial:
if not self._pruned.empty():
kurobako_trial_id, trial = self._pruned.get()
next_step = None
elif self._waitings.empty():
kurobako_trial_id = idg.generate()
trial = self._create_new_trial()
next_step = self._next_step(0)
else:
kurobako_trial_id, trial = self._waitings.get()
frozen_trial = optuna.multi_objective.trial.FrozenMultiObjectiveTrial(
trial._n_objectives,
self._study._storage.get_trial(trial._trial_id))
current_step = frozen_trial.last_step
next_step = self._next_step(current_step)
params = [] # type: List[Optional[float]]
for p in self._problem.params:
if p.is_constraint_satisfied(self._problem.params, params):
params.append(self._suggest(trial, p))
else:
params.append(None)
self._runnings[kurobako_trial_id] = trial
return solver.NextTrial(trial_id=kurobako_trial_id,
params=params,
next_step=next_step)
def ask(self, idg: solver.TrialIdGenerator) -> solver.NextTrial:
params: List[Optional[float]] = []
for p in self._problem.params:
if p.distribution == problem.Distribution.UNIFORM:
params.append(self._rng.uniform(p.range.low, p.range.high))
else:
low = np.log(p.range.low)
high = np.log(p.range.high)
params.append(float(np.exp(self._rng.uniform(low, high))))
trial_id = idg.generate()
next_step = self._problem.last_step
return solver.NextTrial(trial_id, params, next_step)