Zhen-Dong / HAWQ

Quantization library for PyTorch. Support low-precision and mixed-precision quantization, with hardware implementation through TVM.
MIT License
410 stars 83 forks source link

About quantization scheme #33

Open NoLookDefense opened 2 years ago

NoLookDefense commented 2 years ago

Hello. Thanks for your work.I am a newcomer of quantization, and I feel confused about the quantization scheme. It seems that it has so many configurations, like uniform, bops, model size, latency, etc. Could you please explain the differences between these models, and how to train these model? https://github.com/Zhen-Dong/HAWQ/blob/main/model_zoo.md