llama.cpp

be726c57 - gguf.py : uint64_t on all lengths, sizes and counts, enums still uint32_t