Empirical testing (grid search): Jeremy says that in a couple of lessons.
For course 2018 - Take a look at @hiromi post on that:
The rule of thumb for determining the embedding size is the cardinality size divided by 2, but no bigger than 50.
https://medium.com/@hiromi_suenaga/deep-learning-2-part-1-lesson-4-2048a26d58aa