https://github.com/turboderp/exllamav2
https://colab.research.google.com/drive/1yrq4XBlxiA0fALtMoT2dwiACVc77PHou?usp=sharing

2023
https://towardsdatascience.com/exllamav2-the-fastest-library-to-run-llms-32aeda294d26
    https://nuancesprog.ru/p/19534/
    https://vk.com/@nuancesprog-exllamav2-samaya-bystraya-biblioteka-dlya-raboty-s-llm