3 ms·
Llama LORAs are very customizable, and range from "almost full finetuning" to "barely taking more VRAM than GPTQ inference"
by brucethemoose2 3y ago
Llama LORAs are very customizable, and range from "almost full finetuning" to "barely taking more VRAM than GPTQ inference"