mbbl icon indicating copy to clipboard operation
mbbl copied to clipboard

Off-by-one error in `gym_pendulum`

Open bamos opened this issue 4 years ago • 0 comments

Hi, just a minor issue here. The gym_pendulum environment in here runs for 201 steps instead of 200 steps as the original gym pendulum does and I think it's because this line should be an inequality.

Code to see this issue

#!/usr/bin/env python3

import gym
from mbbl.env.env_register import make_env

env = gym.make('Pendulum-v0')
env.reset()

done = False
t = 0
while not done:
    obs, reward, done, _ = env.step(env.action_space.sample())
    t += 1

print(f'Gym Pendulum-v0: {t} steps')

env, _ = make_env('gym_pendulum', rand_seed=0)
env.reset()

done = False
t = 0
while not done:
    obs, reward, done, _ = env.step(env._env.action_space.sample())
    t += 1

print(f'mbbl pendulum: {t} steps')

Output

Gym Pendulum-v0: 200 steps
mbbl pendulum: 201 steps

bamos avatar Jan 26 '20 03:01 bamos