matrix07012 @matrix@gameliberty.club

@r000t So I guess my initial guess was correct, it was spilling over to RAM and PCIe bandwidth became the bottleneck. It just confused me because task manager didn't show any spill over and inference with llama cpp was faster when i let it spill

@teto probably

@PanzerTan you meant arouse, right?

@LukeAlmighty 3080 12gb, i'm using heretic to uncensor InternVL3.5-14B

Why the fuck is it 2x faster on a CPU?

It predicted 50h for small finetune I give up

@nihilvt remaining fedi threads:

@ageha wow, they must be barely selling anything

@ageha buy up their stock and sell it for profit

@sun oven as in for baking or as in for heating (afaik it's both)

I wish I had more VRAM, it's just a 14B parameter model