Best practice for tweaking different architecures in taking the TensorFlow Practitioner exam?

Apologies if this is not an appropriate question for this forum. All background information contained here about the exam is high level and publically available.

I’ve taken the TensorFlow Practitioner exam once, and didn’t make it through. Going back over the questions I’ve worked out where I need to improve my skills.

I’m not very familiar with pycharm. The questions are presented and submitted via a pycharm plugin. I need to test different model architectures in answering questions but I’m worried I will introduce bugs as I tweak the code and be unable to revert to a “clean” test environment.

What is best practice for tweaking code in pycharm without “polluting” the test environment?