llama.cpp

66756c82 - convert-llama-h5-to-gguf.py : add tensor data layout