Bare-metal AI stack for AMD Strix Halo. 109 tok/s. Zero containers. 115GB GPU. Compiled from source.
ai amd arch-linux self-hosted autonomous gaia watchdog bare-metal homelab lemonade rocm ai-agents unified-memory no-cloud llm llama-cpp strix-halo gfx1151 compiled-from-source
-
Updated
Mar 30, 2026 - Python