Cuda memory

[wangyueft]

Cuda memory

[zxczrx123]

@shersoni610 I also had the same problem.
My environment:
Win10(I had changed some code so I can use it in Win10),
One 1080Ti,
Anaconda py3.6,Cuda 9.0,CUDNN 7.5,Pytorch1.1

I solved this problem by set num_workers=0 of DataLoader() in pytorch/main.py .And I also had try to smaller the batch size of train.But in the end I still use 32 and it works.

[shersoni610]

Hello, I tried changing num_workers = 0, but do not know how to change the batch size. The code still fails on 6GB Titan card. Thanks

…

On Mon, Nov 25, 2019 at 12:49 AM zxczrx123 ***@***.***> wrote: @shersoni610 <https://github.com/shersoni610> I also had the same problem. My environment: Win10(I had changed some code so I can use it in Win10), One 1080Ti, Anaconda py3.6,Cuda 9.0,CUDNN 7.5,Pytorch1.1 I solved this problem by set *num_workers=0* of DataLoader() in pytorch/main.py .And I also had try to smaller the batch size of train.But in the end I still use 32 and it works. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#44?email_source=notifications&email_token=ANZR6GTPX6ZPZO4TIBR67CDQVOGSDA5CNFSM4JQ5E7LKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEFBTMGI#issuecomment-558052889>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ANZR6GRE2OXQEK7FMJWU67TQVOGSDANCNFSM4JQ5E7LA> .

[nihil39]

Hi, @shersoni610 @zxczrx123

i'm having the same problem on a meager GT 1030 (RuntimeError: CUDA out of memory. Tried to allocate 80.00 MiB (GPU 0; 1.95 GiB total capacity; 947.23 MiB already allocated; 24.25 MiB free; 1.02 GiB reserved in total by PyTorch)

Changing the number of workers does not help. By the way, what's the point on setting them to zero?

Any help with changing the batch size?

Thank you

[Racketycomic]

Setting the default value of test_batch_size argument in main.py from 16 to 8 worked for me