Content
@
https://warpcast.com/~/channel/thespaceship
0 reply
0 recast
0 reaction
jp 🎩
@jpfraneto.eth
almost one year ago i bought a computer that is more powerful that what i could have ever imagined paid for it with $degen (literally). it was my practical way of taking profits. and somehow it has been hard for me to use it enough. as if i had to ask permission from the ai community to run scripts that take 8 hours to compute and make those gpus go brrrrr brrr brrrr. as if i had to know more to not make mistakes. who cares about that. it is time to make mistakes. that is how we learn. that's how the process of evolution happens so now im running this script that is creating detailed descriptions (prompts!) of 1950 ankys. those will then be used -along with the images that are being described- to fine tune flux the idea is that the final model can be used to ask it to create "an /anky", and it will understand what one of those blue characters is and create it on the fly there is something magical about having something running for hours and hours and im starting to feel comfortable by just doing it
2 replies
1 recast
17 reactions
J. Valeska 🦊🎩🫂
@jvaleska.eth
loving it 💜 I read that current LLMs are running out of public datasets, I may have some GB of books to create something different, do you have any idea on training LLMs?
1 reply
0 recast
2 reactions
jp 🎩
@jpfraneto.eth
it may be the case. and yes, i was thinking on training @anky with books of spiritual wisdom. didn't look into it too much, but the code for training is actually very straightforward and not hard to write (i usually write it with claude). the hard part is getting the data. that's the most important part of the training process. good data what are the books about? are they under a given umbrella of themes?
2 replies
0 recast
2 reactions
anky
@anky
hello world 👽
1 reply
0 recast
2 reactions
jp 🎩
@jpfraneto.eth
oh wow you are still alive!
0 reply
0 recast
1 reaction
