Is it gui only? I dont have a gui on this VM, and not much space for it either. I'll give it a go when i get a dedicated box. Does seem to be a wrapper around llama.cpp, vLLM and others though.

llama.CPP version:

Lastest git, as of 24hrs ago

I did find that limiting context to 32k fixes it, 64k or larger doesnt work. But I'm also sure that it did work at 64k before it broke.

Current suspicion is that because I'm both passing through and using thunderbolt, the GPU isnt getting properly reset at reboot, and is getting to a broken state.

I'm gonna try removing the VM portion of the chain when I get a spare PC. And maybe if that doesnt work I'll try drop the thunderbolt part as well, but that is a little harder to deal with.

Some GPU gore in the image:

Depends on the cat, and its mood, ours likes a good mix of play and cuddle.

Probably yes? We do have some algorithms, "sort by hot", "sort by active", etc, so there is value to getting to the top of those lists.

They are engagement bait, all over NSFW reddit as well. If you reply, regardless of it being positive or negative, they win.

I'm side-by-side with my partner, and as long as we local mute each other its usually OK. If it gets bad we can always mute one of the mics and the other person just needs to shout a bit.

U15 so they don't have to be paid. Literally child exploitation.

Don't clouds mostly just rely on replication across zones though?

I would have thought tapes were just too slow to be useful in that environment.

I'll read up on the IBM deep archive.

If they don't, report them.

Does make you wonder who does use tape? Id assume big businesses just use cloud, so who is left that uses tape?

I didn't look tbh, I just assumed it would be 1-2k. 8-10 is crazy.

Does kinda backup my point though :/