Getting cuda error when upgraded to pytorch 0.4.0

I just updated to 0.4.0 and go the error when running my script:

RuntimeError: cuda runtime error (35) : CUDA driver version is insufficient for CUDA runtime version at torch/csrc/cuda/Module.cpp:51

I’ve tried googling the error but it didn’t help much…Im using conda. Cuda version:

Cuda compilation tools, release 8.0, V8.0.44

I’m happy to provide with more help but not sure what else to provide…

You should update your nvidia driver (not CUDA, the driver)

Hi, I have this problem also. But my setting is different. I have CUDA 7.5 but with Intel GPU instead. I run with Pytorch with CPU only but still not work. My error message is posted here. Thank you in advance.

I don’t think you will be able to use an Intel GPU with CUDA. As far as I know, CUDA is unfortunately only compatible with NVIDIA GPUs.

But I specified the CPU for Pytorch, why does it check my CUDA? It is rather unreasonable.

Have you installed the Cuda-version of pytorch? If you don’t have a suitable GPU you should install the CPU-only version

Thank you very much for your kind reply. I tried to uninstall GPU version and install CPU version of Pytorch. However, the error message now is “AttributeError: module ‘torch._C’ has no attribute ‘_cuda_getDevice’”. I am wondering is it possible that the pre-trained model I am using based on GPU?

Could you show your code (especially the line in which the error occurs)?