My personal advice
If you learn one model and understand it’s concept then shifting from one to another is not tough. It is a few days of work. My advice to you is that you should learn both. But you need not to learn both in-depth. You should start with one and code your models in that framework but you should have knowledge of another framework also. Because it will help you to read other’s codes which are in another framework. You should never be bound by your framework. Though both are good. I started with Keras. And now I am using PyTorch for my work. Because PyTorch is good for experimental works. I liked the pythonic style of the PyTorch. Start with anyone which you think suits you and then try to learn another one also. And if you find another one more comfortable then swift to that. Because most of the core concepts are similar in PyTorch and Keras. The transition from PyTorch to Keras or Keras to PyTorch is easy.
Best of luck for your Deep Learning journey. Keep your focus more on concepts and how they are used and implemented in the real world.