As I have read a little about this I have been thinking I might want to add an AI GPU to my PC. The one I am considering is the NVIDIA GeForce RTX 3060 12GB.
I have some related questions:
Are open source drivers available through Fedora?
Is having this installed compatible with Fedora Workstation?
Is this compatible with the RamaLama Rag project software?
Thanks for your help on this!
Tablepc
I have an RTX 3060 installed on my Fedora Workstation (F43). The Nvidia drivers packaged by the team at rpmfusion work well (not open source). Rock solid performance and stability.
Ramalama rpms are available but not not always current with the latest release.
The instructions at https://copr.fedorainfracloud.org/coprs/g/ai-ml/nvidia-container-toolkit/ are very useful in configuring the cuda toolkit which ramalama can work with.
Be sure to run sudo nvidia-ctk cdi generate --output=/etc/cdi/nvidia.yaml after updating ramalama, the container-toolkit, or the nvidia drivers from rpmfusion.
Not sure what process is relabelling a URL to Making sure you are not a bot. Could not easily find the fix. Hover to get the url.
If you don’t supply an explicit text description of the link, Discourse tries to load the target web page to get its title.
Often this now fails due to AI scraper protection.
The solution is to provide an explicit text description like this:
[our favourite distro](https://fedoraproject.org/)
Discourse makes this easier for you if you click the “link” icon at the top of your post:
You have to balance the benefits of Nvidia and AI with the downsides of Nvidia and Linux.
Normally we don’t recommend Nvidia as they are not fully open. A lot of the ‘problems’ we help solve on this site are Nvidia related.
If you think that Nvidia offers better AI support with ramalama, then maybe you want to accept the non-open nature and install the closed codecs. It’s really your call in this use case.
I fixed it. There have been " in the url. As you used a description and a link text, both had to be removed.
