Make token counting faster and more robust once https://github.com/abetlen/llama-cpp-python/issues/1763 is fixed.
Make token counting faster and more robust once abetlen/llama-cpp-python#1763 is fixed.