TypeError: __init__() missing 1 required positional argument: 'distribution'

I tried to use Ray Tune with with tfp.NoUTurn Sampler but I got this error TypeError: __init__() missing 1 required positional argument: 'distribution'. I tried it with HMC and VI and got the same error. It seems tfp can't find Ray tune's values for the specified hyperparameter.

params = {'num_burnin_steps': tune.randint(30, 100)}

def train model(config):
    #preceding code here...
    num_burnin_steps =config["num_burnin_steps"]
    sampler = tfp.mcmc.TransformedTransitionKernel(
        tfp.mcmc.NoUTurnSampler(
            target_log_prob_fn=pinned_model.unnormalized_log_prob,
            step_size=0.1),
        bijector=constraining_bijector)

   remaining code here...

return {'score': score}

import ray tune and define your parameters

trainable_with_resources = tune.with_resources(train_model, {"cpu": 1})
tuner = tune.Tuner(
    trainable_with_resources,
    param_space=trial_space,
    tune_config=tune.TuneConfig(num_samples=10)
)
results = tuner.fit()`

Tensorflow probability version: 0.19.0
Ray Tune version: 2.3.1

Start by reproducing the shown training snippet with TensorFlow Probability 0.19.0 and Ray Tune 2.3.1, then trace how the Tune configuration reaches the TFP sampler constructors. The issue is done when the source of the missing argument is identified and a verified correction or compatibility explanation is provided.