Does the BayesOptSearch algorithm support discrete hyperparameter spaces in Ray Tune?

Question

I am writing an example using Ray Tune with BayesOptSearch algorithm. However, I got an error at line "results = tuner.fit()" presents

BayesOpt does not support parameters of type Categorical

I am running Ray 2.5.1 on Ubuntu. The problem may be the use of discrete parameters in Bayesian algorithms as blow:

{"param": tune.choice([20, 30])}

The complete code is as follows:

from ray import tune
from ray.air import session
from ray.tune.search.bayesopt import BayesOptSearch


def train_fn(config):
    session.report({"loss": config["param"]})


tuner = tune.Tuner(
    train_fn,
    tune_config=tune.TuneConfig(
        num_samples=100,
        metric="loss",
        mode="min",
        search_alg=BayesOptSearch(),
    ),
    param_space={"param": tune.choice([20, 30])},
)
results = tuner.fit()

I got an answer at

https://github.com/bayesian-optimization/BayesianOptimization/blob/d34f890cfbe4125b1da76790970c3838e9fd5a21/examples/advanced-tour.ipynb

However, I don't know how to use it in Ray Tune.

------------EDIT------------------

When I only switch BayesOptSearch to OptunaSearch which uses default bayesian optimization, the result is OK. The code as follow:

from ray import tune
from ray.air import session
from ray.tune.search.bayesopt import BayesOptSearch
from ray.tune.search.optuna import OptunaSearch


def train_fn(config):
    session.report({"loss": config["param"]})


tuner = tune.Tuner(
    train_fn,
    tune_config=tune.TuneConfig(
        num_samples=10,
        metric="loss",
        mode="min",
        # OptunaSearch which uses default bayesian optimization, OK
        search_alg=OptunaSearch(),
        # BayesOptSearch optimization, ERROR
        # search_alg=BayesOptSearch(),
    ),
    param_space={"param": tune.choice([20, 30])},
)
results = tuner.fit()

best_result = results.get_best_result("loss", "min", "last")
print(f"Best result config: {best_result.config}")