I see a lot of comparison of llm.c with pytorch. But how does it compare against ggml, which is a C++ based library? I think ggml at the moment is mainly used f

I see a lot of comparison of llm.c with pytorch. But how does it compare against ggml, which is a C++ based library? I think ggml at the moment is mainly used for inference and not a lot for training, but still comparing to something equivalent would give us a better understanding of where it stands.