upvote
use llama.cpp with cuda
reply
The problem may be that it's a 7800XT which handles memory contention by freezing.
reply