[Chapter 13] Using tf.data.Dataset with tf.keras


After creating the dataset which prefetches one batch at a time when asked for, How does it exactly integrate with `model.fit()`?
For each iteration/step, does it call the `csv_reader_dataset()` if it does wouldn't it be like sampling without replacement which we don't want and when the author mentioned
> Note that we do not need to repeat it.

Does he mean repeat as in `no_of_epochs` ?