When I have been worked on buckwheat, I used 30 seeds and 5 repetitions. So 150 seeds as a minumum for one test of seed germination (or for one test of early growth).
For maintaning or keeping an optimum population size of a self-pollinated crop (if it is uniform and stable/DUS), you just keep small viable seed population size comparing to that for a cross-pollinated crop (especially for a case of low naturally mutation rate; however, when I worked with inducing mutation, e.g. by irradiation, I tried to keep a large population size during segregation for selection of a desired plant for further evaluation).