Understanding vertical profiles of atmospheric turbulence characteristics is one of the most important problems for theoretical and applied research in the fields of atmospheric optics. Stochastic parallel gradient descent (SPGD) algorithm is proposed for turbulence profile mode fitting. Based on the generalized Hufnagel-Valley model, atmospheric turbulence profile models for different seasons and time of day in Hefei have been developed to fit each observed average C2N vertical profile by SPGD algorithm. The results show that, not only the obtained turbulence mode show best accordance with the observed average profiles of C2N over the whole atmosphere, but also the optical turbulence characteristic parameters of the obtained turbulence modes are in good agreement with those for the average profiles of C2N. The investigation is a useful exploration for developing a “universal method” for turbulence profile model fitting based on the generalized Hufnagel-Valley model.