Hacker News Reader

Comment by minimaxir | Hacker News Reader

Looking at the docs, it appears that the CUDA drivers have to be manually installed on the host image, which takes time/money.

Is there a GCP Image with CUDA drivers pre-installed? Or is that not possible with the hardware architecture?

View on HN

SIGH. Because GPL you can't pre-install the NVIDIA driver. At that point, it sadly makes more sense to have you roll your own and then bake the image. I'll gladly hand out refunds for the X minutes this takes (don't forget that on kernel upgrades you get to do it again when the kernel headers change!), but like you I wish it weren't so.

Feed the installer the right flags and it will run unattended. install Dkms and it'll rebuild automatically on kernel updates

Yes, that's correct. But, you still must have the user do so. Otherwise, you're distributing the resulting artifact (the Nvidia driver linked against the kernel) which is GPL (kernel) and not (NVIDIA).

But IANAL.

Amazon Linux AMI for their GPU instances also does not include the CUDA driver.

Just run a cuda container.

Can you point to an example of this? I have been installing nvidia-docker, and it sounds like what you're proposing is substantially simpler.

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

Edit: still requires nvidia-docker, or a hand crafted docker command that replicates nvidia-docker.

Yeah my use case is nVidia docker but still a lot easier than configuring it on your host system imo

Takes 10 minutes to install nv-docker correctly, and you're golden.

They have a Deep Learning AMI which does, though: https://aws.amazon.com/marketplace/pp/B01M0AXXQB

Reply on news.ycombinator.com