Convergence of Continuous Normalizing Flows for Learning Probability Distributions