A lightweight on-device learning framework for mobile edge devices.
$ git clone https://github.com/FromSystem/Swift-On-Device-Leanringpytorch 1.4.0
torchvision 0.5.0
python 3.7
numpy
logging
time
matplotlib
collections
config.py: All the parameters are set in this file. Please note the logger filepath and data filepath
quantizer.py: Qcon2d and QLinear
shufflenet.py : model network
Timer.py: logger class to log the information when training
shufflenet_qat.py: the main file for fake quantizaiton
data: datasets are stored in this file
./ShuffleNet_QAT/logger: this is a place where you can find out the running infomation.
python shufflenet_qat.py