If you want to understand current LLM craze, check out this talk by Andrej Karpathy:

https://build.microsoft.com/en-US/sessions/db3f4859-cd30-4445-a0cd-553c3304f8e2

[converting and quantizing a model in the background while watching the video]