Tencent/ncnn ? reverse-engineered prompt

Build me a simple working demo around ncnn that can take a small PyTorch model, convert it with pnnx, and run inference from a tiny C++ app. I want the project to load the exported model files, run a sample input, and print the output so I can see the whole flow end to end.

Please make it easy to build on my machine, and include a clear example I can run right away. If Vulkan support is available, wire it in so it can use the GPU when possible, but keep a CPU fallback too. I’d also like a small Python example if that’s straightforward, so I can test the model before using the C++ version.

Keep it clean and beginner friendly, with brief setup steps and a sample model or toy network included. If you need to look up the current build or export instructions online, go ahead and do that.