I’m also not a mentor for MLS, but as Balaji says the issue is probably “versionitis”. The APIs in the python/ML space are not that stable over time. These courses were published several years ago using the versions that were current at that time. You can’t just port the notebooks to your local environment with whatever current versions you happen to have and assume everything will “just work”.
You have a fundamental choice of two approaches to this kind of issue:
- Just debug each instance you run into.
- Set up an environment that allows you to duplicate the particular versions to match a given notebook.
Approach 2) is a lot more work, but once you have that understanding then it’s a general tool that you can use a lot over time. Using approach 1) is less work in any given case, but it’s not predictable how many such issues you have to fight your way through each time you try to port something.
Here’s a thread which will get you started on the process of figuring out option 2), but it is beyond the scope of these courses.