f0ff8 - 2y
So initial impression of #LLAMA 13B is it's really really cool especially for an #LLM that runs off a laptop from 10GB of training data, it's actually super impressive this has been developed so rapidly, but it is alpha software... Def worth playing around with.
33093 - 2y
Wow this looks really cool. Didn’t know you can run this locally 😳 Questions. Does it censor? Do you have to jailbreak it to get it to say interesting stuff? Whats the generation speed like? Whats your hw specs like if you don’t mind sharing.
No censorship. No need to jailbreak. There's even already forks you can load with nsfw training data. For the main project look up llama.ccp. It uses leaked data from Facebook's ChatGPT competitor. For the fork that runs Hugging Face data (including nsfw) look up bloomz.ccp. Give it a month and this will be a bigger game changer than ChatGPT.
Holy shitballs. I know what to do this weekend haha. Thanks for the info man.
Oh and my hardware specs:
Haha have fun! It's seriously amazing to be on the cutting edge of this shit.
