mrdbourke/pytorch-deep-learning — reverse-engineered prompt

Build me a beginner friendly PyTorch deep learning course repo that feels like a complete learn by doing curriculum, not just random notes. I want it centered around notebooks with clear explanations, runnable code, simple visuals, and a few helper utilities so someone can go from zero to building real models.

Start with PyTorch basics, then a practical model building workflow, classification, computer vision, working with custom datasets, turning notebook code into cleaner reusable scripts, transfer learning, tracking experiments, trying to replicate a research style result, and finally a simple model deployment example. Include sample data, saved models, demo assets, and docs or a book style guide so the lessons are easy to follow in order.

Please make it feel polished for self study, with exercises or practice prompts where it makes sense, and keep the teaching style very hands on. If anything is unclear, check the current PyTorch docs and common learning patterns online and fill in the gaps with sensible defaults.