deSilva

If you are a #freebsd user who has been wanting to try #ollama, I have opened a PR open to add support for the platform

https://github.com/ollama/ollama/pull/14697

Add FreeBSD Support by Aesthetikx · Pull Request #14697 · ollama/ollama

These changes were sufficient to get Ollama running successfully for me on FreeBSD, utilizing either GPU via Vulkan or CPU inference. Compared to #4172, I think these changes are simpler, as the pr...

GitHub
Mar 7 at 7:23pmWeb
Show threadbenMar 7
@aesthetikx personally i would consider non-support of ollama to be a positive feature of freebsd, but that's just me
Show threaddeSilvaMar 7
@benjamineskola I suppose I can't disagree with you. I am also concerned to see that the project now supports 'cloud' models, rather than just local inference.
Show threadbenMar 7
@aesthetikx Is local inference actually any better? Seems to me like it has most of the same problems.
Show threadEnigmaRotorMar 7

@aesthetikx While I am quite against AI on the “AI Bubble/gafam level, and all the social consequences, I can see some uses at a local, individual level, and offline initiatives like ollama. They can be useful and give some draft to a clean human brain work. Would be great to have them usable from FreeBSD.

(But I would rage against anyone using AI to commit some code in the FreeBSD project. Because we don’t deserve shit to be inserted in the tree. Shit on the branches of a tree and you might expect gravity to return the present to yourself)

Show threaddeSilvaMar 7
@EnigmaRotor Agreed. My use case is to summarize and tag bookmarks for local indexing, which even a low-grade open weight LLM is sufficient for.
Show threadᴏᴏᴍ-ᴋɪʟʟᴇʀ: 333Mar 7

@aesthetikx

cc: @feld