Examples of NVIDIA accelerated libraries in action. These libraries make up the CUDA layer of AI's 5-layer cake (energy, chips, infrastructure, models, applications), turning chips into usable infrastructure for the models and applications stacked on top. Each subfolder demonstrates a single library through hands-on usage and a real-world workflow.
Copy TEMPLATE/ and rename it to your library (e.g. RAPIDS-cuDF/), then open a pull request. Only modify files inside your own folder.
A contribution typically covers:
- Purpose & setup — what the library does and how to install it.
- Basic usage — a working demo of core functionality.
- First Bowl of Soup — A real world workflow where the library shines, and how it fits end-to-end.