Squashed Gaussian Sample Seeded

Problem

Sample u ~ Normal(mu, sigma) with rng = random.Random(seed), then squash it with tanh. Return [u, tanh(u)] so callers can also use the pre-squash value in the log-probability.

Hint

SAC samples via reparameterization so the policy stays differentiable.

Test cases

case 1

in: [0,1,42]

expected: [-0.14409032957792836,-0.1431013398880218]

case 2

in: [1,0.5,0]

expected: [1.4708577023403322,0.8997409430816226]

case 3

in: [-1,2,7]

expected: [-1.5117605768952007,-0.907250987286807]

Starter codePython

import random, math
def squashed_gaussian_sample(mu, sigma, seed):
    # Your code here
    pass

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