How difficult it is to fine tune model like this with specific domain knowledge?... | Hacker News

Hacker News new | past | comments | ask | show | jobs | submit

login

iamcreasy on Sept 27, 2023 | parent | context | favorite | on: Mistral 7B

How difficult it is to fine tune model like this with specific domain knowledge? I am currently looking into gpt-3.5-turbo-instruct for this same purpose.

brucethemoose2 on Sept 28, 2023 [–]

The hardest part is formatting the data correctly. Garbage in garbage out.

But in terms of hardware, it should be very cheap and doable on most 8GB+ Nvidia GPUs, and maybe AMD/Intel GPUs on linux.

Consider applying for YC's Spring batch! Applications are open till Feb 11.
Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact