Repository navigation
One error #75
Description
RuntimeError: CUDA out of memory. Tried to allocate 1024.00 MiB (GPU 0; 7.94 GiB total capacity; 7.03 GiB already allocated; 144.81 MiB free; 7.12 GiB reserved in total by PyTorch)
I meet this question after i print "python exps/cifar_train" in Terminal ,and it apears after
"path:logs/cifar_train_2022_03_22_19_29_36
0%| |0/1563 [00:00<?, it/s]" .
I know this means the CUDA is out of memory, but i only run this one program ,and the image has not loaded. Did author or someone also meet this question and how did you deal with it?
Activity
yifanjiang19 commentedon Mar 30, 2022
what's your batch size and what's the number of gpu you used?
Nanboy-Ronan commentedon Apr 10, 2022
I have the same problem with only one GPU.
yifanjiang19 commentedon Apr 14, 2022
@Nanboy-Ronan What's your batch size?
Nanboy-Ronan commentedon Apr 14, 2022
Thank you for reply. My training batch_size is 64. Probably I should adjust it? But I see there is people saying that they train with 8 gpus for 1.5 days, which made me hesitate in continue doing this.
wudiduojimone commentedon Apr 14, 2022
Thanks for your patient reply, and I'm sorry for I saw it just now. Maybe I also have the problem like Nanboy-Ronan for I only have one GPU.
yifanjiang19 commentedon Apr 17, 2022
@Nanboy-Ronan one gpu is not able to train it. But if you still one to train it, you should tune your batch size to a much smaller size.
KapparapuVenkataAjithKumarReddy commentedon Apr 22, 2022
May i know how to do this this was my major project.