How to set predefined trials?

[vl-dud]

Let's say I have initial assumptions regarding optimal hyperparameters. So I use the code:

import optuna

sampler = optuna.samplers.TPESampler(seed=42)
study = optuna.create_study(direction="minimize", sampler=sampler)
distributions = {
	"x": optuna.distributions.IntDistribution(1, 10),
	"y": optuna.distributions.IntDistribution(1, 10)
}
predefined_params = [
	{
		"x": 1,
		"y": 2
	},
	{
		"x": 3,
		"y": 4
	}
]
for params in predefined_params:
	trial = optuna.trial.create_trial(
		state=optuna.trial.TrialState.WAITING,
		params=params,
		distributions={k: v for k, v in distributions.items() if k in params}
	)
	study.add_trial(trial)
	
def objective(trial):
	# Some objective
	
study.optimize(objective, n_trials=100)

to add two predefined trials to the study. I suggest that two first trials will be defined by me, and the rest will be random from the search space. Will this work correctly? Did I choose a right way to solve the problem? Maybe it would be better to customize the sampler or use FixedTrial objects?

[nzw0301]

Study.enqueue_trial is all you need.

[vl-dud]

Thank you!