GPU memory overflow when execute reseting the environment for multiple times

[fujiawei0724]

When running the following commands:

import gym

env = gym.make('NeedlePick-v0')

for _ in range(100000000):

env.reset()

My GPU occupied memory will gradually overflow. Would you please help me solve the problem? Thank you very much

[TaoHuang13]

Hi Jiawei,

Thanks for your question. I have run the commands with 10K iterations, and it seems no overflow is found (mine is 24G memory capacity).

If you are running DEX or ViSkill, I remembered that 4G memory will be enough to handle 100K iterations without overflow problems.

So would you please try to use a GPU with more memory, or provide more details to see through that deeply?

[fujiawei0724]

Thanks for your prompt answer. It seems like the pybullet version cause this problem. I want to know your pybullet version when you run the code, since the surrol only instructs the pybullet >= 3.0.7, which is not a specific one. I guess the problem will be solved if I hold the same version of pybullet with you. Thank you very much. Looking forward to your reply~

[fujiawei0724]

I have figure out the problem, the memory leak is caused by the high version of pybullet. The version 3.0.9 has solved the problem, but in the following version, the problem reappear. In a word, the pybullet version should >= 3.0.7 and <= 3.0.9. Thanks for your help again.