llama.cpp GPU Usage Failure Due to Incorrect Backend

This occurs when using a wrapper or library, such as llama-cpp-python, that is not compiled with the necessary GPU support flags. You can diagnose this by checking the installation logs or metadata of the library to see if it was compiled with GPU-specific CMake arguments. If the package was installed via a standard pip command without environment variables like CMAKE_ARGS, it likely defaulted to a CPU-only build.

Sources