because I use a high performance clustering and its highest cuda version is 11.6, and I cannot install cuda myself. So I want to ask if I can install pytorch2.0 under cuda version 11.6?
Yes, since the PyTorch binaries ship with their own CUDA libraries and you locally installed CUDA 11.6 toolkit will only be used if you build PyTorch from source or custom CUDA extensions.
thank you. could you give me some example that how I build pytorch from custom CUDA exension?
You can take a look at this tutorial but note that it’s not necessary to build and use custom extensions.
What I meant in my previous post is that your locally installed CUDA toolkit will not be used to execute any plain PyTorch code, but only if you build PyTorch from source or a custom CUDA extension (which is optional).