R2D2 Burn-In

Problem

Compute the number of usable steps after R2D2-style burn-in: seq_len - burn_in, floored at 0 when the burn-in consumes the whole sequence.

Hint

Burn-in lets the recurrent state warm up before gradients are applied.

case 1

in: [80,40]

expected: 40

case 2

in: [40,40]

expected: 0

case 3

in: [20,40]

expected: 0

case 4

in: [10,0]

expected: 10

Starter codePython

def r2d2_usable_steps(seq_len, burn_in):
    # Your code here
    pass

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