search

Sentdex / pygta5

Explorations of Using Python to play Grand Theft Auto 5.
MIT License
3.91k stars 823 forks source link

Resource exhausted ERROR #98

Closed hemanth2410 closed 2 years ago

hemanth2410 commented 6 years ago

while training with Tensorflow GPU i am getting resource exhausted error error

hemanth2410 commented 6 years ago

PC specs Intel core i7 7700HQ @ 3.4 GHz 8 GB SODIMM DDR 4 RAM GTX 1050 4GB(NETBOOK)

DevJhns commented 6 years ago

if it's not problem in path to files, you can try add batch_size=32 or 16 or even 4 (by default its 64) to model.fit() . Neural networks are really heavy for gpu memory, so 16 worked for me, problem is, it may cause some acc and loss issues on later epochs, but without this can't run at all, so i'm curious does it affects a lot, I mean, will model be okay with such size, or its bad idea.

kunakl07 commented 5 years ago

What is the command that you used for training ?

Denrose11 commented 5 years ago

if it's not problem in path to files, you can try add batch_size=32 or 16 or even 4 (by default its 64) to model.fit() . Neural networks are really heavy for gpu memory, so 16 worked for me, problem is, it may cause some acc and loss issues on later epochs, but without this can't run at all, so i'm curious does it affects a lot, I mean, will model be okay with such size, or its bad idea.

how to decrease the batch size in model.fit() for version .03