This one log line was triggering a single line llama.log to be generated in the pwd of the server
This changes the model for llama.cpp inclusion so we're not applying a patch, but instead have the C++ code directly in the ollama tree, which should make it easier to refine and update over time.