Python Forum

Full Version: Why Cuda from numba package is unavailable?
You're currently viewing a stripped down version of our content. View the full version with proper formatting.
Python 3.8.5, numpy 1.19.2, numba 0.51.2, Kubuntu 20.04, Nvidia-driver 390.138

I have video- card Nvidia GTX 560 Ti and I can't use one in calculations:
from numba import cuda
is_available = cuda.is_available()
print(str(is_available))
Output:
False

   
detect = cuda.detect()
print(str(detect))
Output:
Found 1 CUDA devices id 0 b'GeForce GTX 560 Ti' [SUPPORTED] compute capability: 2.1 pci device id: 0 pci bus id: 1 Summary: 1/1 devices are supported True
My GPU is supported, because has capability 2.1: supported devices. And I successfully installed cuda- toolkit with:
$ conda install cudatoolkit

Why?
It turned out, that it was necessary to run the python interpreter, which was installed together with the anaconda-package (in the home directory). In this interpreter, cuda is supported.
I install cuda- toolkit, and error disappear.