[enlyth]

Exactly, we're just below that sweet spot right now.

For example on 24GB, Llama 30B runs only in 4bit mode and very slowly, but I can imagine a RLHF finetuned 30B or 65B version running in at least 8bit would be actually useful, and you could run it on your own computer easily.

[bick_nyers]

Do you know where the cutoff is? Does 32GB VRAM give us 30B int8 with/without a RLHF layer? I don't think 5090 is going to go straight to 48GB, I'm thinking either 32 or 40GB (if not 24GB).

[riku_iki]

> For example on 24GB, Llama 30B runs only in 4bit mode and very slowly

why do you think adding vram, but not cores will make it run faster?..

[enlyth]

I've been told the 4 bit quantization slows it down, but don't quote me on this since I was unable to benchmark at 8 bit locally

In any case, you're right it might not be as significant, however, the quality of the output increases with 8/16bit, and running 65B is completely impossible on 24GB

[riku_iki]

It's not impossible, there are several projects which load model layer by layer for execution from the disk or ram, but it will be much slower.