NVIDIA/TensorRT ? reverse-engineered prompt

Build me a working TensorRT OSS project that I can clone, set up, and run on an NVIDIA GPU machine.

I want the repo to build the open source pieces of TensorRT, including the sample apps and the ONNX model import flow, so I can test that inference works end to end. Please make it easy to get started with a clean install, and add whatever scripts or clear instructions are needed so someone can build it without getting stuck.

If there are any current docs or version requirements I should follow, look them up online and use the latest compatible setup. I’d also like a simple way to run at least one sample after the build so I can confirm everything is working.