Cuda return to false

Hi I’m new with programming.

I’m facing some problem with cuda installation. I want to know, is cuda toolkit installation
a. in window - using downloaded installer from (
b. in python - usingconda install pytorch torchvision torchaudio cudatoolkit=8.0 -c pytorch
the same thing?
Do I need both installed or only in python?

Hope someone can help me with this. Thanks.

If you install the binaries (via conda or pip) they will ship with their own CUDA runtime and you would only need to install the corresponding NVIDIA driver on your system.
Have a look here (Table 1) to find the appropriate driver versions for different CUDA releases.

I’m using window 10. I have GPU nvidia quadro 6000 version 377.83 (latest). If I refer to this link Check GPU-Cuda Computibility and quadro 6000 computibilily is 2.0 which match with cuda 8.0

However, still torch.cuda.is_available() returning to False. Is this a common problem?

PyTorch supports CUDA>=9.2 and the binaries ship with a min. compute capability of 3.7, so your device won’t be supported anymore.

I see. That solves the problem. No wonder the “false” problem keeps repeating.

Thanks a lot!