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