Wait, the ZimaCube has a private GPT implementation?

MigratingtoLemmy@lemmy.world to Selfhosted@lemmy.world – 17 points –

Hi everyone, was just going over the ZimaCube page on ZimaBoard's website: https://www.zimaboard.com/zimacube/

They mention a "private GPT" for one's files: I'm interested in how they implemented it, if the process is open source, and if self-hosting something like this is possible.

Thanks!

5

100% it's just an half assed hack around an open source library

For Kickstarter it doesn't need to actually work, it just needs to have the option in the menu

Probably running a local model, like for example Mistral. Sounds like small local model + RAG

"GPT" describes a machine learning technique - tools like ChatGPT use this technique along with massive training sets to produce their results, but there isn't anything stopping you doing it on a small scale so you can slap "GPT" or "AI" on your product and jump on the band wagon without actually adding anything of value

There are tons of options for running LLMs locally nowadays, though none come close to GPT4 or Claude 2 etc. One place to start is /c/localllama@sh.itjust.works