What are you talking about? Fine tunes are basically just more of the same train... | Hacker News

Hacker Newsnew | past | comments | ask | show | jobs | submit

		hnfong on Jan 1, 2024 \| parent \| context \| favorite \| on: Stuff we figured out about AI in 2023 What are you talking about? Fine tunes are basically just more of the same training, optionally on selected layers for efficiency.

imjonse on Jan 2, 2024 [–]

RLHF or DPO are definitely not just the same thing as the basic torch training loop, hence my many more lines of code argument.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact