CUDA runtime only installation

[leondavi]

Hi all,
I have searched for installation of CUDA API and CUDA Driver runtime and couldn’t find it in Nvidia’s website. The regular toolkit installation contains development tools and I search for without any development tools such as Nsight or GDB, only the minimum that is required for running CUDA compiled applications.

Where can I find CUDA runtime only packages (RBM for centos 6)?

[Robert_Crovella]

NVIDIA doesn’t publish instructions for this formally that I know of. However NVIDIA does maintain reference docker images for 3 different levels of CUDA support. You could refer to the dockerfiles for some guidance if you like:

[url]Docker Hub

For example the base CUDA 10.1 Centos 6 dockerfile is here:

[url]https://gitlab.com/nvidia/cuda/blob/centos6/10.1/base/Dockerfile[/url]

[leondavi]

@Robert_Crovella

NVIDIA doesn’t publish instructions for this formally that I know of. However NVIDIA does maintain reference docker images for 3 different levels of CUDA support. You could refer to the dockerfiles for some guidance if you like:

https://hub.docker.com/r/nvidia/cuda

For example the base CUDA 10.1 Centos 6 dockerfile is here:

https://gitlab.com/nvidia/cuda/blob/centos6/10.1/base/Dockerfile

Thank you for your answer. Those links are helpful.