My 16 year old nephew lives in an East African nation where there is practically no internet access.
Last week he asked me for advise as to how to go about training an open source LLM using an on disk Wikipedia (~80 GB).
Any suggestions? Thanks!
My 16 year old nephew lives in an East African nation where there is practically no internet access.
Last week he asked me for advise as to how to go about training an open source LLM using an on disk Wikipedia (~80 GB).
Any suggestions? Thanks!
5 comments
Here's his "1 hour intro to LLMs" video: https://www.youtube.com/watch?v=zjkBMFhNj_g
1. https://www.youtube.com/c/AndrejKarpathy
I will try to download it and send it to him.
Most decent LLMs probably were already trained on wikipedia, that doesn't stop them from hallucinating when asked questions about it.
I will do some research over the weekend. Thanks for mentioning it!
You can download llamafile and several models, put them on a USB drive or hard drive, them send the drive to him via DHL.
I think he wants to tinker, and learn more about how they work. What I neglected to mention is that he's already learned to program (developing Android apps, and he's also learned Python). He is a very bright and curious kid.
LLM training in simple, raw C/CUDA
----------------------------------
https://github.com/karpathy/llm.c
It is only 1,000 lines of easy to read C code. There is also Python reference code.
The project is an idea at the moment. My contact in Kenya has direct access to the Principals of the schools that our supported students attend.
My thought is that the teachers would not have to do much. Many of the students already know python and could do self-learning individually or in groups.
A flash drive with llamafile+models and documentation might be all that it would take to get them started - even offline.
Bonus: Using llamafile, the same binary distribution works on MacOS, Linux, and Windows.
I wasn't aware of Phi-3 - I will look into it.