llama.cpp
convert : force patch_embd weights to F16 or F32 to avoid broken GGUFs
#15367
Merged
Merged