In some recent papers, a lot have come up with a quantization on 2 bits and 4 bits. How can I do that on tensorflow? As far as I know, 8-bits is the least now.
