Closed VODKA312 closed 1 year ago
I am reproducing this model. But it happend many errors. I just don't know how to figure it out.
Sorry for my stupid question. The reason is my cuda cnn is unmatched about my graphic card. I changed another graphic card that is 2080Ti and cuda version is 10. It works. So if other people want to reproduce this code. My suggestions is use older than 20x graphics card.
My software and hardware configuration: ubuntu 22.04 CUDA 11.8 graphics card 4060Ti docker and docker -gpu- container, the docker image is 'upsnet'. you can find it in docker hub.
I try to run it. But I found my graphics card and OS cannot support pytorch 1.0.0 and pytorch 0.4.1. So I use docker image to solve this problem. In docker image, the cuda version is 10, and the torch.is_avilable() and nvidia-smi is ok. and Its implement 'init.sh' in UPSNET sucessfully. But, When i try to test the modle, many errros are throw out, the detailed informtion is list here:
I just want to know how to solve this problem. I am exhaused about it. I think it might be some wrong in my grahpics card or gpu or torch_version.