News for hackers. Stuff that matters.

A comment on LLaMA2 isn't “Open Source” – and why it doesn't matter. Parent comment

Re:LLaMA2 isn't “Open Source” – and why it doesn't matter by lolinder on Friday July 21, 2023 @03:21PM (#36818287)
Not that it's okay for this to be in the license, but I'm curious: what is the use case for synthetic data? Most of the discussion I've seen has been about how to avoid accidentally using LLM-generated data.
[ Parent | on HN ]
Re:LLaMA2 isn't “Open Source” – and why it doesn't matter by lmeyerov on Friday July 21, 2023 @03:23PM (#36818313)
Tuning a tiny classifier
[ Parent | on HN ]
Re:LLaMA2 isn't “Open Source” – and why it doesn't matter by int_19h on Friday July 21, 2023 @08:30PM (#36821908)
You can use synthetic data produced by more complex models to finetune smaller ones to be better.
[ Parent | on HN ]