Open ZFTurbo opened 4 years ago
I trained XNORNet with your code and checked weights after training. I expected XNORNetConv2D to have -1 or 1 values, but it has float values like: 0.05971663, 0.07685672 etc. While BinaryNetConv2D has weights as expected -1 or 1. What can be wrong?
I trained XNORNet with your code and checked weights after training. I expected XNORNetConv2D to have -1 or 1 values, but it has float values like: 0.05971663, 0.07685672 etc. While BinaryNetConv2D has weights as expected -1 or 1. What can be wrong?