We conduct hydrodynamical MOND simulations of isolated disc galaxies over the stellar mass range M_⋆/M_⊙ = 10⁷ - 10¹¹ using the adaptive mesh refinement code PHANTOM OF RAMSES (POR), an adaptation of the RAMSES code with a Milgromian gravity solver. The scale lengths and gas fractions are based on observed galaxies, and the simulations are run for 5 Gyr. The main aim is to see whether existing sub-grid physics prescriptions for star formation and stellar feedback reproduce the observed main sequence, and Kennicutt-Schmidt relation that captures how the local and global star formation rates relate to other properties. Star formation in the models starts soon after initialisation and continues as the models evolve. The initialized galaxies indeed evolve to a state which is on the observed main sequence, and the Kennicutt-Schmidt relation. The available formulation of sub-grid physics is therefore adequate and leads to galaxies that largely behave like observed galaxies, grow in radius, and have flat rotation curves - provided we use Milgromian gravitation. Furthermore, the strength of the bars tends to be inversely correlated with the stellar mass of the galaxy, whereas the bar length strongly correlates with the stellar mass. Irrespective of the mass, the bar pattern speed stays constant with time, indicating that dynamical friction does not affect the bar dynamics. The models demonstrate Renzo's rule and form structures at large radii, much as in real galaxies. In this framework, baryonic physics is thus sufficiently understood to not pose major uncertainties in our modelling of global galaxy properties.