As parameters for train_task and eval_task, respectively, how can we live without wrapping the data generators with itertools.cycle like we did from exercises in previous weeks?
Hi @vhah
In loose words, the train_task and eval_task have internal itertools.cycle usage. You can imagine the train_task receiving the iterator traing_generator which it can call whenever it wants (and it knows that the train_generator has methods like next and other that it would ever need).
Cheers