Not everyone! but seems like a lot of people are moving towards PyTorch. But isn’t that the case for any technology? someone has to start and then we create something better or different.
In my opinion, the major difference between TensorFlow and PyTorch is the way computational graphs are represented. Static (TensorFlow) vs. dynamic graph representations (PyTorch)…
In Tensorflow, you have to pre define the computation graph of your model. But in PyTorch, you can define or even manipulate the graph at any time. This is particularly helpful while using variable length inputs in RNNs.
One great benefit of TensorFlow is the boards (TensorBoard). It lets you visualise the ML models within the browser.
It comes down to the taste. TensorFlow still have a much bigger community behind it and who is to say that it wont evolve past this point.
Both TensorFlow and PyTorch have contributed immensely to the community.