def log_prob_t(self, state, action):
        return self.distribution_t(state).log_prob(action)[:, None]