ShuffleNet Channel Shuffle Index

Problem

Compute the channel shuffle permutation used by ShuffleNet.

Reshape the channel axis to (groups, channels // groups), transpose to (channels // groups, groups) and flatten; return the resulting index list. groups divides channels.

Test cases

case 1

in: [4,2]

expected: [0,2,1,3]

case 2

in: [6,3]

expected: [0,3,1,4,2,5]

case 3

in: [4,1]

expected: [0,1,2,3]

case 4

in: [8,4]

expected: [0,4,1,5,2,6,3,7]

Starter codePython

def channel_shuffle_indices(channels, groups):
    # Your code here
    pass

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