News for hackers. Stuff that matters.
Re:Mixtral 8x22B by htrp on Wednesday April 17, 2024 @11:01AM (#40065667)
maybe we'll look back at token context windows like we look back at how much ram we have in a system.
Re:Mixtral 8x22B by frabjoused on Wednesday April 17, 2024 @11:04AM (#40065709)
I agree with this in the sense that once you have enough, you stop caring about the metric.
Re:Mixtral 8x22B by paradite on Wednesday April 17, 2024 @11:07AM (#40065763)
And how much RAM do you need to run Mixtral 8*22B? Probably not enough on a personal laptop.
Re:Mixtral 8x22B by user_7832 on Wednesday April 17, 2024 @11:13AM (#40065857)
Generally about ~1gb ram per billion parameters. I've run a 30b model (vicuna) on my 32gb laptop (but it was slow).
Re:Mixtral 8x22B by Lacerda69 on Wednesday April 17, 2024 @11:12AM (#40065839)
I run it fine on my 64gb RAM beast.
Re:Mixtral 8x22B by coder543 on Wednesday April 17, 2024 @11:16AM (#40065887)
At what quantization? 4-bit is 80GB. Less than 4-bit is rarely good enough at this point.
Re:Mixtral 8x22B by apexalpha on Wednesday April 17, 2024 @11:18AM (#40065927)
Is that normal ram of GPU ram?
Re:Mixtral 8x22B by samus on Wednesday April 17, 2024 @04:22PM (#40069671)
64GB is not GPU RAM, but system RAM. Consumer GPUs have 24GB at most, those with good value/price have way less. Current generation workstation GPUs are unaffordable; used can be found on ebay for a reasonable price, but they are quite slow. DDR5 RAM might be a better investment.
Re:Mixtral 8x22B by htrp on Wednesday April 17, 2024 @11:47AM (#40066338)
While there is a lot more HBM (or UMA if you're a Mac system) you need to run these LLM models, my overarching point is that at this point most systems don't have RAM constraints for most of the software you need to run and as a result, RAM becomes less of a selling point except in very specialized instances like graphic design or 3D rendering work.

If we have cheap billion token context windows, 99% of your use cases aren't going to hit anywhere close to that limit and as a result, your models will "just run"

Re:Mixtral 8x22B by bamboozled on Wednesday April 17, 2024 @11:20AM (#40065958)
I still don’t have enough RAM though ?
Re:Mixtral 8x22B by samus on Wednesday April 17, 2024 @04:29PM (#40069745)
RAM is simply too useful.