News for hackers. Stuff that matters.
Re:The False Promise of Imitating Proprietary LLMs by brucethemoose2 on Thursday May 25, 2023 @11:30PM (#36079467)
The jump between llama 13B and 30B is quite significant. And their instruction finetuning is not SOTA I don't think, though the point about general knowledge is a good one: instruction llama lies very confidently.

But one great thing about open source LLMs is that you can specialize them in various tasks with affordable LORA training, enough to easily beat GPT4 in a specific niche.

Re:The False Promise of Imitating Proprietary LLMs by ThorsBane on Friday May 26, 2023 @08:31PM (#36090992)
Any recommended starting points for LORA training llama 30B on a specific niche? Books, tutorials, videos are all appreciated. Thanks for your time!
Re:The False Promise of Imitating Proprietary LLMs by f_devd on Friday May 26, 2023 @09:23PM (#36091233)
Currently SOTA for specialization of LLMs is QLoRA: https://github.com/artidoro/qlora
Re:The False Promise of Imitating Proprietary LLMs by ThorsBane on Friday May 26, 2023 @09:29PM (#36091253)
Thank you very much my friend. Be blessed in all that you do. <3