llama.cpp

convert : write Gemma embedding scale for DFlash drafts

#29802

Merged