NVIDIA/DeepLearningExamples ? reverse-engineered prompt

Build me a polished repository of deep learning examples that shows how to train and run a bunch of popular models on NVIDIA GPUs, with a strong focus on speed, reproducible accuracy, and easy deployment.

I want it organized by model type, like computer vision, NLP, speech, recommendation, forecasting, and a few other common workloads, with clear training and inference workflows, example notebooks, and enough docs that someone can get started without reading a giant guide. Use the main frameworks people expect for this kind of project, and make sure the examples are set up to work well with NVIDIA optimized containers and GPU acceleration. Include support for mixed precision and multi GPU where it makes sense, and add simple paths for exporting or serving models when that’s relevant.

Keep the structure clean and practical, with a README that explains what each example does, how to run it, and what hardware it’s meant for. If you need current framework or container details, look them up online first.