HER Future Goal Pick

Problem

Sample future goals for hindsight relabeling with a seeded random generator.

Draw k indices uniformly from [t + 1, T - 1] using rng.randrange(t + 1, T). Return [] when there is no future timestep.

Hint

The future strategy relabels with states that were actually reached later.

case 1

in: [0,5,3,42]

expected: [1,1,3]

case 2

in: [2,6,4,0]

expected: [4,4,3,4]

case 3

in: [3,4,5,7]

expected: []

case 4

in: [1,3,2,123]

expected: [2,2]

Starter codePython

import random
def her_future_goals(t, T, k, seed):
    # Your code here
    pass

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