simonko

lets hope that it will work on gpus like the rx 5700 xt what has limited rocM support

is there anything you have always wanted to achieve? that idea you know you want to be true? is possible if any

keep working 💪

achieve maximum benchmark score per param

lol

you're talking about robust performance and not benchmaxxing, right? right?

atleast i dont do benchmaxxing lol

is there anything you have always wanted to achieve? that idea you know you want to be true? is possible if any

keep working 💪

i had 2 goals:
make the smartest and smallest model
make the most hilarious model

so yeah... lol

im alredy trying kinda simular stuff but im bad at finding ok datasets, and i usualy do 0.2-0.5b and 4k ctx with llama tokenizer

cool, mind sharing it?

i have some in my profile, not the best for now but im testing random ideas yeah, around largest i make are 0.5b

yeah im making some