okuvshynov / slowllama

Finetune llama2-70b and codellama on MacBook Air without quantization
MIT License
448 stars 34 forks source link

use conf instead of magic 4 #13

Closed roryclear closed 9 months ago