Temperature Scaling on Logits

Problem

Apply temperature scaling and return softmax(logits / temperature) as a list of probabilities. Use the max-subtraction trick for numerical stability. Return an empty list when logits is empty or temperature <= 0.

Test cases

case 1

in: [[1,2,3],1]

expected: [0.09003057317038046,0.24472847105479764,0.6652409557748218]

case 2

in: [[0,0],2]

expected: [0.5,0.5]

case 3

in: [[1,1],2]

expected: [0.5,0.5]

case 4

in: [[],1]

expected: []

case 5

in: [[2],0]

expected: []

Starter codePython

import math
def temperature_scale(logits, temperature):
    # Your code here
    pass

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