Description:

We currently provide a Python demo for Phi4 in the LLM-TPU project:

LLM-TPU repository: https://github.com/sophgo/LLM-TPU/tree/main
Phi4 model demo (Python): https://github.com/sophgo/LLM-TPU/tree/main/models/Phi-3

Task:

Implement a C++ version of the Phi-4 chat/demo that achieves functionality and output quality comparable to the existing Python demo. There is an example in Qwen3, you can also refer to it.

Requirements:

Ensure generated responses are similar in style and quality to the Python demo.