A 70B GGUF model stalls on an AMD R9700 as VRAM hits 30 GB but RAM stays flat

Developer-Y · reddit · 2026-07-28

A Reddit user is trying to run AstroSage 70B on an AMD Radeon AI Pro R9700 with ROCm 7.2 and llama.cpp, but the model stalls while loading.

Original post →

More from Infra

Infra channel →