I am trying to understand how to use QNNPACK. The repo lacks documentation (as it is already mentioned in issues raised in github repo). Can some one explain how to use it? In the repo desciption it is mentioned " As of today, QNNPACK is integrated in PyTorch 1.0 with Caffe2 graph representation." but no where in the Pytorch documentation explains how it is integrated and more importantly, what are the hooks/funtions are availabe for Pytorch users to use this library.
I am also interested in how to use qnnpack. All I found in sources, that pytorch jit can quantize linear layers.
But I’m thinking it is applicable for amd64 processors, not for arm… Trying to find some tests in sources to quantize model, but no luck