[am17an]

One often overlooked after that is ggml, the tensor library that runs llama.cpp is not based on pytorch, rather just plain cpp. In a world where pytorch dominates, it shows that alternatives are possible and are worthy to be pursued.