Gent is right that the current version of the assignment doesn’t have that misleading comment. Are you taking the course through your university or some other source besides directly from Coursera? Why would you have the older version?
The larger point is that we always use the from_logits = True mode for efficiency. That means the sigmoid (or softmax) is applied in the loss calculation. Here’s a thread which discusses that. And here’s a thread from mentor Raymond that shows why that approach is better.