LocalLLaMA
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.
Rules:
Rule 1 - No harassment or personal character attacks of community members. I.E no namecalling, no generalizing entire groups of people that make up our community, no baseless personal insults.
Rule 2 - No comparing artificial intelligence/machine learning models to cryptocurrency. I.E no comparing the usefulness of models to that of NFTs, no comparing the resource usage required to train a model is anything close to maintaining a blockchain/ mining for crypto, no implying its just a fad/bubble that will leave people with nothing of value when it burst.
Rule 3 - No comparing artificial intelligence/machine learning to simple text prediction algorithms. I.E statements such as "llms are basically just simple text predictions like what your phone keyboard autocorrect uses, and they're still using the same algorithms since <over 10 years ago>.
Rule 4 - No implying that models are devoid of purpose or potential for enriching peoples lives.
view the rest of the comments
If you alternatively want to spend LESS time keeping up to date on them, just keep up with what Unsloth offers quants for because they'll have most of the interesting models anyway, and they'll immediately list how much RAM you need for whichever quant on their Docs page
Actually…
And I really don’t mean to be pedantic. But Unsloth has enshittified some.
Some of their GGUFs target their own vibecoded PRs first, not necessarily mainline llama.cpp.
And, more frequently as of recent times, they’re bugged just to get them out the door first, and get the hype.
The quantization loss is higher than some other quanters now, again, due to the rush.
There are also some much stronger quantization schemes outside mainline GGUFs now, like ik llama.cpp or exl3.
Unsloth isn’t bad, I’m just saying be careful using them as the defacto quantization source.
I used to love them. But I haven’t downloaded an Unsloth model in some time, and TBH I am not really inclined to…
Llama.cpp themself is slow, lot of model lose their popularity.
Yeah, hence the slew of forks.
Not that I blame llama.cpp. They have a huge job for so little dev manpower, and it’s trying to be a jack of all trades rather than something specialized like other frameworks.
Heh, good call. I'm interested in the theory, but I could just dig into them as Unsloth surfaces them. Worth thinking about.