The flake.lock is so old that newer ROCm GPUs are not supported, for instance. https://github.com/ggml-org/llama.cpp/blob/33983057d0f578aca74ba15eccc3de9c267a5ff6/flake.lock#L21 @ngxson I think it is your domain. Or do you want me to submit a PR?
