Terrasque
- 0 Posts
- 165 Comments
I mean, I totally agree with you. But that also kinda ignores all the useful things a dog can be trained to do.
It’s less the calculations and more about memory bandwidth. To generate a token you need to go through all the model data, and that’s usually many many gigabytes. So the time it takes to read through in memory is usually longer than the compute time. GPUs have gb’s of RAM that’s many times faster than the CPU’s ram, which is the main reason it’s faster for llm’s.
Most tpu’s don’t have much ram, and especially cheap ones.
Reasonable smart… that works preferably be a 70b model, but maybe phi3-14b or llama3 8b could work. They’re rather impressive for their size.
For just the model, if one of the small ones work, you probably need 6+ gb VRAM. If 70b you need roughly 40gb.
And then for the context. Most models are optimized for around 4k to 8k tokens. One word is roughly 3-4 tokens. The VRAM needed for the context varies a bit, but is not trivial. For 4k I’d say right half a gig to a gig of VRAM.
As you go higher context size the VRAM requirement for that start to eclipse the model VRAM cost, and you will need specialized models to handle that big context without going off the rails.
So no, you’re not loading all the notes directly, and you won’t have a smart model.
For your hardware and use case… try phi3-mini with a RAG system as a start.
22·6 months agoI’m not saying it’s broken, but it has some design choices and functions that makes even Whatsapp a better choice for privacy minded people. Like rolling their own crypto and not having e2ee as default.
4·6 months agoKoboldcpp is way easier. Download exe, double click exe, open gguf file with the AI model, click start.
Then put on your robe and wizard hat
7·6 months agoSo you’re saying it’s already feature complete with most json libraries out there?
59·6 months agoYou realise there is no algorithm behind Lemmy, right?
Of course there is. Even “sort by newest” is an algorithm, and the default view is more complicated than that.
You aren’t being shoved controversial polarizing content subliminally here.
Neither are you on TikTok, unless you actively go looking for it
4·7 months agoI’ve seen Skype do that. It was a weird folder name, but gallery found it and displayed the images.
Which is how I noticed it in the first place
I wonder what bpm Moby’s thousand starts at… Maybe it can reach both limits
8·8 months agoWho are you?
What do you want?
Also, I think good and bad is a bit fluid there. It’s just people with different agendas. Well, except emperor Cartagia. And perhaps Bester.
Hah as if. In the early 00s the mods were in maybe once or twice a day and there was tons of CP being posted.
Worst I saw was a little girl chopped into pieces, and a many -page discussion / argument if it should be sorted as CP or Necro porn. That was the old 4chan.
Even for 4chan that’s fucked up.
Oh, sweet summer child…
Yep, I usually make docker environments for cuda workloads because of these things. Much more reliable
They had a trial run with bleach already
On occasion their strategy has been “if we send in enough people, they’ll eventually run out of bullets”
They out-Zapp Brannigan’ed Zapp Brannigan. That should terrify you on multiple levels
Koboldcpp
Like when under Arab spring the Egyptian politicians tried to get the military involved to stop the protests, and got back (paraphrased)
“Our primary job is to protect the Egyptian people from violence. You really don’t want us involved in this”