SAME Padding Sizes

Problem

Compute the asymmetric padding needed for SAME padding with a stride. The output length is ceil(n / stride); total padding is max(0, (output - 1) * stride + dilation * (k - 1) + 1 - n), split as floor(total / 2) before and the remainder after.

The signature is same_padding(n, k, stride=1, dilation=1). Return [pad_before, pad_after].

Hint

Extra padding on odd totals goes after the input, matching common framework behavior.

Test cases

case 1

in: [5,3]

expected: [1,1]

case 2

in: [6,3,2]

expected: [0,1]

case 3

in: [7,3,2]

expected: [1,1]

case 4

in: [4,3,3]

expected: [1,1]

case 5

in: [5,3,1,2]

expected: [2,2]

Starter codePython

def same_padding(n, k, stride=1, dilation=1):
    # Your code here
    pass

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