llama.cpp

Remove split metadata when quantize model shards

#6591

Merged