If you have enough regular ram and an nvme ssd you can run huge models on modest hardware using Colibrí.
I got a Lenovo mini PC with 64GB of ram and a 500GB nvme and a 38B model runs with 11 tokens per second on my machine.
The 38B model fits in my ram but if I had used a 120B parameter model then it would stream the rest of the model from my nvme disk.
Gyrcas@piefed.zip 2 days ago
Please don’t run LLMs at all…
db2@lemmy.world 2 days ago
They have uses, most of which aren’t what’s being pushed right now and none of which are what the tech bro sociopaths are pushing.
jpablo68@infosec.pub 1 day ago
I mainly use it to practice german.