Tag
A pull request (PR#27210) for adaptive MTP has been submitted to llama.cpp, a C/C++ implementation for LLM inference with minimal setup and high performance.