We introduced some cool LLM sampling algorithms and their implementation on GPUs.