Ask your advisor – it's literally their job to give you advice! In particular, they know your situation much better than we do, so they can give much better advice about how much time you should be spending on reviews.
A general rule of thumb is that you should review at least as many papers as you submit.* You're doing plenty more than that, I assume, so it's absolutely reasonable to decline review requests if they're taking up too much of your time: for example if they're taking up too much of your research time or leaving you with too little free time to relax. And, of course, this will vary through time. You might decline a review request that comes in when you're really busy with your own research (maybe a deadline's coming up or you're visiting some other researcher) even though you'd have accepted it if it came a few weeks earlier or later.
Remember, though, that reviewing papers has benefits to you. It keeps you in tune with developments in your field and exposes you to new ideas. Writing good reviews also makes more senior people in your field (the people you write reviews for) more aware of your talents.
* This will vary from field to field and depend on how many authors and how many reviewers a paper typically has. In my area of computer science, 2–4 authors and 2–4 reviews is about normal, so a one-to-one ratio is reasonable.