I disagree in the case of numpy. If you're going to be using python for numerical work almost all of your data will be numpy arrays so there is not reason to not get used to them from the beginning. You don't have to go digging into the obscure corners of numpy, but you should be familiar with basic arrays and how they work.
It's not a given that you'll need to work with numpy arrays. It's quite likely that numpy arrays will be used behind the scenes, but you'll be working in a different abstraction (e.g. pandas dataframes or tensorflow tensors) that will hide pretty much all of numpy stuff from you.
But I'd argue that even in those cases they are easier to understand if you already know numpy arrays as they share a lot of syntax. And if you're ever going to interact with other python numeric libraries you're sooner or later going to be converting your dataframes and tensors to and from numpy arrays anyway.
