Hi,
where to start if I want to add support for a new parallel computing device to pytorch?
Found it here Extending dispatcher for a new backend in C++ — PyTorch Tutorials 1.10.1+cu102 documentation