User avatar
Two Hollywood Phonies @autumn@cafe.autumn.town
1y
you can get LLMs in 0.5 bil parameters, and I've finetuned BERT in the past (110M parameters), with ~10k lines of data taking 40 minutes. so i wonder if i could finetune an actual chatbot LLM for the sake of posting like me
1
0
1
0
User avatar
Two Hollywood Phonies @autumn@cafe.autumn.town
1y
I'd need to filter the data somehow though because a lot of it is kinda crap
1
0
2
0
User avatar
no longer kopper @kopper@wetdry.world
1y
@autumn i was kinda curious about this as well but i only have an 8gb rx580 on windows where the only compute api that works with LLMs is fucking vulkan
1
0
2
0
User avatar
Two Hollywood Phonies @autumn@cafe.autumn.town
1y
@kopper ill will figure this out with my not great but better 12gb 6700XT
1
0
2
0
User avatar
outdated undefined ✨ @twilight@labyrinth.zone
1y
@autumn @kopper realising just how bad nvidia is, 2060 has 6gb
2
0
1
0
User avatar
no longer kopper @kopper@wetdry.world
1y
@twilight @autumn yeouch!!
0
0
2
0
User avatar
Two Hollywood Phonies @autumn@cafe.autumn.town
1y
@twilight @kopper the 4060 has 8GB 😭, and the 4060 Ti has a 8gb and (a lot more expensive) 16GB option
:Hollow@wetdry.world:1
0
0
1
1