car_racing.py memory leaking issue

[jackyoung96]

gym/gym/envs/box2d/car_racing.py

Line 527 in 38a1f63

vl = pyglet.graphics.vertex_list(

There is memory leaking in car_racing.py code.

pyglet.graphics.vertex_list() function makes the memory almost double,

And it can be fixed with vl.delete() under this line

gym/gym/envs/box2d/car_racing.py

Line 530 in 38a1f63

vl.draw(gl.GL_QUADS)

[praveenVnktsh]

Thanks a lot. This fixed the leak issue.

@Jaekyung-Cho I guess it would be great if you could submit a pull request for this.

[JunkyByte]

Hello, I've experienced the same memory leak and applied the solution given by @Jaekyung-Cho.
While it reduces the problem (greatly) it does not eliminate it, I'm running out of ram pretty quickly using multiple CarRacing environments in parallel (on a 64gb ram machine).

Could you share your results for the following code? I see constant increase in ram usage.
(For fair comparison I'm sharing a single environment test code to reproduce so that there's no doubt that it is related to gym / CarRacing env).

import gym
import psutil
import os

def memory_used():
    return psutil.Process(os.getpid()).memory_info().rss * 1e-6  # To megabyte

e = gym.make('CarRacing-v0')
e.unwrapped.verbose = 0

for i in range(10000000):
    e.reset()

    if i % 1000 == 0:
        print('Ram Used: %f' % memory_used())

which outputs (numbers are megabyte)

Ram Used: 153.264128
Ram Used: 172.523520
Ram Used: 181.149696
Ram Used: 181.149696
Ram Used: 197.632000
Ram Used: 197.632000
Ram Used: 197.632000
Ram Used: 202.498048
Ram Used: 204.931072
Ram Used: 226.336768
Ram Used: 229.040128
Ram Used: 231.743488
Ram Used: 234.176512
Ram Used: 236.609536
Ram Used: 239.312896
Ram Used: 241.745920
Ram Used: 244.449280
Ram Used: 280.604672
Ram Used: 283.037696
(...)

looking for a solution, would you be so kind to tell me if you are able to reproduce the issue?

[jackyoung96]

Hello, I've experienced the same memory leak and applied the solution given by @Jaekyung-Cho.
While it reduces the problem (greatly) it does not eliminate it, I'm running out of ram pretty quickly using multiple CarRacing environments in parallel (on a 64gb ram machine).

Could you share your results for the following code? I see constant increase in ram usage.
(For fair comparison I'm sharing a single environment test code to reproduce so that there's no doubt that it is related to gym / CarRacing env).

import gym
import psutil
import os

def memory_used():
    return psutil.Process(os.getpid()).memory_info().rss * 1e-6  # To megabyte

e = gym.make('CarRacing-v0')
e.unwrapped.verbose = 0

for i in range(10000000):
    e.reset()

    if i % 1000 == 0:
        print('Ram Used: %f' % memory_used())

which outputs (numbers are megabyte)

Ram Used: 153.264128
Ram Used: 172.523520
Ram Used: 181.149696
Ram Used: 181.149696
Ram Used: 197.632000
Ram Used: 197.632000
Ram Used: 197.632000
Ram Used: 202.498048
Ram Used: 204.931072
Ram Used: 226.336768
Ram Used: 229.040128
Ram Used: 231.743488
Ram Used: 234.176512
Ram Used: 236.609536
Ram Used: 239.312896
Ram Used: 241.745920
Ram Used: 244.449280
Ram Used: 280.604672
Ram Used: 283.037696
(...)

looking for a solution, would you be so kind to tell me if you are able to reproduce the issue?

Ram Used: 173.060096
Ram Used: 181.374976
Ram Used: 181.428224
Ram Used: 196.329472
Ram Used: 194.228224
Ram Used: 196.370432
Ram Used: 198.959104
Ram Used: 201.515008
Ram Used: 218.734592
Ram Used: 221.319168
Ram Used: 223.891456
Ram Used: 226.594816
Ram Used: 229.027840
Ram Used: 231.723008
Ram Used: 234.295296
Ram Used: 236.916736
Ram Used: 268.730368
Ram Used: 271.433728
Ram Used: 273.866752
Ram Used: 276.467712
Ram Used: 279.158784
Ram Used: 281.751552

I also got same result.

In my case, I didn't run car_racing module more than 24 hours. And that's why I didn't care about this RAM memory leaking.

Anyway, thank you for notice me and I try to solve it!!!