ok asking again becase I want to do this . I want to train an LLM on my posts. I'm gonna do it on my own computer. what's a model I can use which doesn't suck major ass

@sierra i think. anything less than 1 bill parameters on ollama will train within a reasonable amount of time. but people just be using markov bots and gpt2 for that stuff

@autumn gpt 2 huh ..

@sierra yeah its pretty shit for actual things but if you just need to make posts you can use that, its a bit more advanced than markov.

idrk how fine-tuning over LLMs would work tho

@sierra probably a good place to start would be at huggingface or ollama docs

@sierra there's ministral (Mistral, european), gemma (da goog), maybe da Qwen's (Alibaba)? idk how good Qwen is anymore within the world of SLMs... but anyways yea these are the 3 big ones :3

Ministral and Gemma is available in 1B so u shouldnt have much problem, idk abt Qwen tho. hope that helps!

@wintburr I was looking at ministral earlier I might use that . problem now becomes How tf do I even train / fine-tune it on my posts and can I do it without needing to use p*thon 🥀

@sierra i lied on accident there is no Ministral 1B and even 3B is commercial-only. fuck. anyways

i don’t think you will be able to avoid Python sadly as the two biggest open-source frameworks to train AI models are PyTorch and TensorFlow, which are both Python…

as for how, I’m looking at it right now, as I haven’t done it since GPT-2, but i’m very curious! :)