How to use the stable-baselines.stable_baselines.common.distributions.MultiCategoricalProbabilityDistribution function in stable-baselines

def __init__(self, nvec, flat):
        """
        Probability distributions from multicategorical input

        :param nvec: ([int]) the sizes of the different categorical inputs
        :param flat: ([float]) the categorical logits input
        """
        self.flat = flat
        self.categoricals = list(map(CategoricalProbabilityDistribution, tf.split(flat, nvec, axis=-1)))
        super(MultiCategoricalProbabilityDistribution, self).__init__()

def proba_distribution_from_flat(self, flat):
        return MultiCategoricalProbabilityDistribution(self.n_vec, flat)

def probability_distribution_class(self):
        return MultiCategoricalProbabilityDistribution