#!/usr/bin/env python3
# -*- coding: utf-8 -*-
import gym
from torch.distributions import Normal
mean, std = 0, 1
# Number of samples to take and size of elite set.
N, n = 100, 10
ITERS = 100
# TODO Apply objective function to X and sort by those indices.
for t in range(ITERS):
X = Normal(mean, std).sample_n(N)
X, _ = X.sort(dim=0,
descending=False) # swap descending to min/max objective
elite = X[:n]
mean, std = elite.mean(), elite.std()
# env = gym.make('CartPole-v0')
env = gym.make("MountainCarContinuous-v0")
s, done = env.reset(), False
while not done:
a = Normal(mean, std).sample()
succ, r, done, _ = env.step(a.numpy())
if __name__ == "__main__":
print(mean, std)
