Weight quantization

Weight quantization is a technique that reduces the number of bits used to represent weights. This is done by converting model weights from high-precision floating-point representations to low-precision floating-point or integer representations