LocalLLaMA

2951 readers

33 users here now

Welcome to LocalLLaMA! Here we discuss running and developing machine learning models at home. Lets explore cutting edge open source neural network technology together.

Get support from the community! Ask questions, share prompts, discuss benchmarks, get hyped at the latest and greatest model releases! Enjoy talking about our awesome hobby.

As ambassadors of the self-hosting machine learning community, we strive to support each other and share our enthusiasm in a positive constructive way.

founded 2 years ago

MODERATORS

SkySyrup@sh.itjust.works

pax@sh.itjust.works

noneabove1182@sh.itjust.works

Smokeydope@lemmy.world

MonsterBug@sh.itjust.works

What is better: higher quantiation or higher parameter count? (yiffit.net)

submitted 2 years ago by Wander@yiffit.net to c/localllama@sh.itjust.works

16 comments fedilink hide all child comments

For example, does a 13B parameter model at 2_K quantiation perform worse than a 7B parameter model at 8bit or 16bit?

you are viewing a single comment's thread
view the rest of the comments

[–] rufus@discuss.tchncs.de 9 points 2 years ago* (last edited 2 years ago) (2 children)

https://github.com/ggerganov/llama.cpp#quantization

https://github.com/ggerganov/llama.cpp/pull/1684

Regarding your question: 13B 2_K seems to be on par with 7B 16bit and 8bit. Not much of a difference between all those. (Look at the perplexity values. Lower is better.) The second link has a nice graph.

Most people don't go as low as 2bit though. It's considerably worse than 4bit.

[–] Wander@yiffit.net 5 points 2 years ago

That graph is great. Very easy to understand. Thank you!

[–] noneabove1182@sh.itjust.works 2 points 2 years ago

These are good sources, to add one more, the GPTQ paper talks a lot about perplexity at several quantization and model sizes:

https://arxiv.org/abs/2210.17323