llama.cpp

2ab0707a - convert : use 'model' value if it exists. This allows karpathy/tinyllamas to load (#4089)