swz30
Seem that the version of pytorch dismatch with the gpu
I run the code on an nvidia gpu A6000, I met the following error:
Traceback (most recent call last): File "demo.py", line 127, in
it seems that the version of torch is too old; if I update the torch version to 2.0.1 with cuda11.8 in the environment , it would failed when I build the basicsr through the following command:
python setup.py develop --no_cuda_ext
the failed reason: ModuleNotFoundError: No module named 'torch'
So when i keep 10.2, CUDA error: no kernel image is available for execution on the device;when i update to 11.8, i can't build the environment.Have you encounter problem like this ? Is the error owing to the old version of torch?
Perhaps you can keep the PyTorch version unchanged while upgrading the CUDA version conda install pytorch==1.8.1 torchvision==0.9.1 torchaudio==0.8.1 cudatoolkit=11.3 -c pytorch -c conda-forge
