• 0 Posts
  • 8 Comments
Joined 8 months ago
cake
Cake day: November 23rd, 2024

help-circle


  • I used the recommended migration tool and it worked okay for many containers but iirc the docker ones had to have one of the security options manually changed in their config which didn’t transform properly with the tool (maybe nesting enable?).

    May very well have changed in the meantime or I only made a mistake, that was in my experimentation phase.

    Ultimately, I did rebuild my instances from the ground since I also switched file system, and to make better use of incus profiles (e.g. one with docker provisioned, one with monitoring and so on) so I couldn’t give you a long-term migration review.

    For me that was (relatively) painless by just migrating the docker volumes in place and rebuilding the stacks, of course ymmv.

    If you decide on migrating and stumble upon issues don’t hesitate to hit me up - I’m only an amateur but maybe I can still help!


  • After having my dinky homelab machine on proxmox for a couple years, since the start of the year I am now running basically everything under a clean Debian system using incus and docker on the individual lxc guests.

    Incus has completely replaced proxmox for me and it’s so much easier to reason about (for me at least) that I wanted to maybe point your cold hands in that direction too ;)



  • One point I would refute here is determinism. AI models are, by default, deterministic. They are made from deterministic parts and “any combination of deterministic components will result in a deterministic system”. Randomness has to be externally injected into e.g. current LLMs to produce ‘non-deterministic’ output.

    There is the notable exception of newer models like ChatGPT4 which seemingly produces non-deterministic outputs (i.e. give it the same sentence and it produces different outputs even with its temperature set to 0) - but my understanding is this is due to floating point number inaccuracies which lead to different token selection and thus a function of our current processor architectures and not inherent in the model itself.