Int8 Conv Output Scale

Problem

Compute the affine quantization parameters for int8 tensors.

scale = (fp_max - fp_min) / 255 and zero_point = round(-fp_min / scale). Return [round(scale, 6), zero_point].

case 1

in: [-1,1]

expected: [0.007843,128]

case 2

in: [0,6]

expected: [0.023529,0]

case 3

in: [-3,3]

expected: [0.023529,128]

case 4

in: [0,255]

expected: [1,0]

Starter codePython

def int8_scale(fp_min, fp_max):
    # Your code here
    pass

4 test cases · run your code against them in the editor