MediumNLPnlp-384

Temperature Softmax Probabilities

Problem

Sampling temperature rescales logits before softmax: p = softmax(logits / T).

Given the logits and temperature T > 0, return the probabilities computed stably.

Hint

Low temperature sharpens, high temperature flattens.

case 1

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

expected: [0.09003057317038046,0.24472847105479764,0.6652409557748218]

case 2

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

expected: [0.1863237232258476,0.3071958857184984,0.506480391055654]

case 3

in: [[0,0],0.5]

expected: [0.5,0.5]

Starter codePython

def temperature_softmax_probabilities(logits, temperature):
    # Your code here
    pass

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