Quantization of a pytorch model

quantization
1

is pytorch QAT pytroch version or cudnn version or Nvidia gpu agnostic?

2

QAT should work on all recent versions of PyTorch and whatever cudnn/gpu versions pytorch supports.