fn architecture(path: &Path) -> Option<String>Expand description
The architecture a gguf declares, read from the file’s own header.
Read BEFORE the load, and that order is the point. llama.cpp loads an
encoder-decoder gguf without complaint and asserts inside the first
decode, and a ggml assert aborts the process - so the file has to be
refused while refusing is still possible. None when the header is not a
GGUF one, is malformed, or declares nothing, which the caller refuses on
the same grounds as an architecture it cannot run.