Boltzmann generators (BGs) [1] are neural network-based importance samplers that can generate configurations of molecular systems. In contrast to classical molecular sampling techniques, BGs generate independent samples from the Boltzmann distribution in one shot…