Docker Compose for Qwen3.6 27B OBLITERATED ROCmFP4 + MTP at five 262K slots on AMD Strix Halo (gfx1151), with optional stock Vulkan and Laguna ROCmFPX stacks.
docker-compose amd vulkan mtp rocm laguna llama-cpp llm-inference gguf openai-compatible strix-halo gfx1151 qwen3-6 rocmfp4 rocmfpx
-
Updated
Jul 26, 2026 - Python