llama.cpp

quantize: skip tensor override when in fallback mode

#14995

Merged