DoReFa-Net in PyTorch

Link to the paper: https://arxiv.org/abs/1606.06160.

Partial implementation supporting 1-bit weights and k-bit activations only.

Contains a single example, on MNIST.

Use with QNN-MO-PYNQ

The scripts trains a LeNet run python qnn_mnist.py --ab k (where k is the activation bits). For use with the Finnthesizer in my QNN-MO-PYNQ Fork, run the python qnn_mnist.py --export to create a compatible NPZ archive.