高通的paper
A White Paper on Neural Network Quantization
https://arxiv.org/pdf/2106.08295.pdf
非对称量化
Q=R / S + Z
R = S(Q - Z)
对称量化
把zero_point固定到0
R = S * Q
Q = R / S
A White Paper on Neural Network Quantization
https://arxiv.org/pdf/2106.08295.pdf
Q=R / S + Z
R = S(Q - Z)
把zero_point固定到0
R = S * Q
Q = R / S