NVIDIA's TensorRT Cookbook Drops v1.0.0 With Consumer GPU Runtime Support
NVIDIA cut the first tagged release of its open-source TensorRT Cookbook, updating all examples to TensorRT 10.16 and adding a new TensorRT-RTX section.
PRO- NVIDIA tagged v1.0.0-trt-10.16 of its TensorRT Cookbook, the first formal versioned release.
- All examples updated to TensorRT 10.16, with cuda-python API changes from the 10.13 update baked in.
- New section 10-TensorRT-RTX covers the consumer-GPU runtime plus API-diff examples versus mainline TensorRT.
- Container matrix maps NGC Docker images to exact TensorRT, CUDA, and driver versions, useful for pinning deploys.
- Covers plugin authoring, ONNX parser integration, multi-stream inference, and trtexec tooling in Python.
- Apache 2.0 licensed, 1.7k stars, still carries some Chinese hackathon material alongside translated English content.
NVIDIA has published the first formally versioned release of its TensorRT Cookbook, tagged v1.0.0-trt-10.16. The repository, which lives under NVIDIA/trt-samples-for-hackathon-cn, has been a rolling collection of TensorRT sample code for years. This release pins everything to TensorRT 10.16, refreshes the API examples, and introduces a dedicated section for TensorRT-RTX, the runtime aimed at consumer GPUs.
The cookbook targets engineers wiring TensorRT into a production inference stack rather than calling it through a high-level wrapper. The earlier jump to TensorRT 10.13 changed cuda-python APIs and broke a lot of code in the wild. The new tag consolidates those fixes into a single reference point.
What is actually in the box
The repo is organized as a numbered curriculum rather than a set of standalone demos. The top-level sections cover:
- 01-SimpleDemo and 03-Workflow: end-to-end examples of building a network and pushing a PyTorch model through ONNX into a TensorRT engine.
This story is for Pro members
You've reached the end of the free preview. Upgrade to AlphaSignal Pro to read the full article - and everything else behind the paywall.
Comments
