I want to understand in which cases and how (automatically, when model is built, or not) PyTotch uses topological sorting.
I can't check it manually, cause of some circumstances, so I've tried searching anything with pytorch and topological sorting key words, and 've found nothing. But I've found that for example TensorFlow definitely does use topological sorting automatically. I expect that PyTorch either does and even in every case (because, gradient back propogation would take exponentially more time to process without topological sorting in common case) but want to make sure.