The old Dockerfile used the deprecated llama.cpp/ subdirectory approach which doesn't exist in ollama v0.23.2. Now using the official CMake presets (ROCm 6 preset) with AMDGPU_TARGETS including gfx906:xnack-.
The old Dockerfile used the deprecated llama.cpp/ subdirectory approach which doesn't exist in ollama v0.23.2. Now using the official CMake presets (ROCm 6 preset) with AMDGPU_TARGETS including gfx906:xnack-.