Comment by colorant

The ipex-llm implementation extends llama.cpp and includes additonal CPU-GPU hybrid optimizations for sparse MoE