llama.cpp

Fix conversion of some BERT embedding models

#6937

Merged