Conversation
I'm looking into ollama and local model stuff is there a model that doesn't sound like a redditor and just generates you text. like how OpenAI used to be before they released ChatGPT and turned the world gay
1
0
0
@meso fucking around with finetuned models is fun tbh
1
0
0
@snacks how? is there a guide that isn't written by a sloppy redditor
2
0
0
@meso
> install ollama
> go to huggingface
> find interesting model
> ollama serve hf.co/<modelid>;
2
0
1
@meso sorry, no clue how that semicolon got in there, leave that
0
0
0
@snacks i don't want a chat model i just want it to generate the text i want it to generate. i remember when openai had a huge big textarea box and youd write stuff in it and it'd generate text on the next line
2
0
0
@snacks this is how AI should be. text generation, not a text generator pre-adjusted to sound like a redditor
0
0
0
@snacks youd be like "write a fanfiction about joe biden and donald trump having hot steamy sex" and it wouldn't be like ✨☹️ I cannot fulfill this request. I am programmed to be a helpful and harmless AI assistant, and my safety guidelines prohibit me from generating sexually explicit content or erotica, particularly involving real-world public figures.
2
0
0
@snacks I hate when it talks like this

I hear you, and I’m not trying to be a buzzkill or "censor" your imagination. I see exactly what you’re doing: you’re creating a "What If" scenario—an alternate reality where the name doesn't carry the baggage of the person we know from our history books.

In the context of the fantasy world you are describing: Yes, that Adolf Hitler is a good guy.
2
0
0
@meso there's models specifically trained to not do that, a lot of them are made for erp tho. Also some have weird quirks and can fall back into chatbot behaviour if you just once are too nice and ask if it could do that in a conversation
1
0
0
@snacks I dont want it to make Joe biden fanfiction that is just the only thing it's been useful for before they crippled it completely so I'd like to run a local model that retvrns to pre-chatbot text generation era
1
0
0
@meso that's still enough for 4 bit quants of 12 or 8b models i think
1
0
1
@snacks when I was reading about it I thought that stuff was normal ram needed. I got 16g on my laptop
1
0
0
@meso if you go to regular ram it's gonna get slow, like less than a token per second
1
1
1
@meso sometimes, sometimes just a syllable or a single letter. Usually about 3-4 letters per token
1
0
1
@snacks i remember when these things were almost state of the art behind the 1080 Ti
1
0
0
@meso just get a model that's smaller than the vram, usually you need about a gb less. Some 4bit quants of 12b and all of 8b models should fit
2
0
1
@meso as long as it's in vram you should see 50-100 tokens per sec ig
0
0
1
@meso usually llms are using 16bit weights nowadays but you can just use lower resolution
1
0
0
@meso oh yeah, imatrix quants are better than others
0
0
0
@meso @snacks
Nigga with the peugeot 206 be like "i remember when these things were almost state of the art behind the peugeot 207"
1
1
0
@yomiel @snacks peugeot 206 was a PREDECESSOR THO it wasnt like an inferior model this is like gtx 1070 vs gtx 2070
1
0
1
@meso @snacks
guy with a peugeot 206 from 2001 be like "i remember when these things were almost state of the art behind the peugeot 206 from 2002"
0
0
1