llama.cpp

model: move `load_hparams` and `load_tensors` to per-model definition

#22004

Merged