Cosine Schedule Value at Step

Problem

Cosine annealing sets lr_t = lr_min + 0.5 * (lr_max - lr_min) * (1 + cos(pi * t / T)).

Given lr_max, lr_min, t, and total steps T > 0, return the learning rate at step t.

Hint

The cosine sweeps from 1 to -1 as t goes from 0 to T.

case 1

in: [0.1,0,0,100]

expected: 0.1

case 2

in: [0.1,0.001,50,100]

expected: 0.0505

case 3

in: [0.2,0,100,100]

expected: 0

Starter codePython

def cosine_schedule_value_at_step(lr_max, lr_min, t, total):
    # Your code here
    pass

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