NVlabs / GA3C

Hybrid CPU/GPU implementation of the A3C algorithm for deep reinforcement learning.
BSD 3-Clause "New" or "Revised" License
649 stars 195 forks source link

Segmentation fault #44

Open cyou6 opened 3 years ago

cyou6 commented 3 years ago

./_train.sh: line 3: 3010 Segmentation fault (core dumped) python GA3C.py "$@"

Does anybody get a segmentation problem like this?

ifrosio commented 3 years ago

Never seen this. What environment are you using? Can you update the drivers?

cyou6 commented 3 years ago

Thanks, I use Ubuntu 16.04, tf 1.14 python 3.7

you mean which driver to update?

On Sun, Feb 7, 2021 at 5:12 PM iuri frosio notifications@github.com wrote:

Never seen this. What environment are you using? Can you update the drivers?

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/NVlabs/GA3C/issues/44#issuecomment-774639888, or unsubscribe https://github.com/notifications/unsubscribe-auth/AFHQXQPS52IRGERV45VVRXLS5ZKRBANCNFSM4XHG23PQ .

ifrosio commented 3 years ago

GPU driver

cyou6 commented 3 years ago

Ok, I will try. Do you mind telling me what does the args "$@" mean in the _train.sh? python GA3C.py "$@"

this line in ga3c.py will return a valueerror since no '=' is given: x, y = sys.argv[i].split('=')

Thanks

On Sun, Feb 7, 2021 at 5:40 PM iuri frosio notifications@github.com wrote:

GPU driver

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/NVlabs/GA3C/issues/44#issuecomment-774643288, or unsubscribe https://github.com/notifications/unsubscribe-auth/AFHQXQNMM5JEAAXYKMJ5VWDS5ZNXNANCNFSM4XHG23PQ .

ifrosio commented 3 years ago

$@ is all of the parameters passed to the script in the sh file. Please also double check that your config file contains correct parameters. It may be that your GPU does not have enough RAM - what GPU are you using? You can try directly modifying the script parameters when calling train.sh - see the docs.

cyou6 commented 3 years ago

ah, I am using 2080. It might be the ram problem since I'm running a couple of training tasks as the same time. I will try again once those tasks are done. Thanks