// the find
jeffheaton/app_deep_learning
T81-558: PyTorch - Applications of Deep Neural Networks, Washington University in St. Louis
This is the notebook set for Jeff Heaton's T81-558 deep learning course at Washington University in St. Louis, covering PyTorch from basic tensors through CNNs, LSTMs/transformers, GANs, explainability, and RL. It's for someone who wants a full semester's worth of structured, code-first deep learning material rather than a library or tool.
Wide and current syllabus — it covers classic architectures alongside 2026-relevant topics like embeddings, diffusion models, and calling API-based LLMs, all with runnable PyTorch notebooks instead of slides. It's actively maintained against a real, running course (last push matches the Fall 2026 semester), so the notebooks get re-verified against current library versions each term rather than rotting. Datasets are provided as direct downloads, so you're not stuck scraping Kaggle or digging through broken links.
This is course scaffolding, not a library: assignment notebooks are literally named assignment_yourname_*.ipynb and tied to specific due dates and in-person meeting weeks, so a chunk of the repo is irrelevant if you're not enrolled. No tests or CI — if a notebook breaks against a newer PyTorch or pandas release mid-semester, you won't know until someone hits it manually. Datasets are hosted on the instructor's personal S3 bucket (data.heatonresearch.com) rather than versioned with the repo, so there's no guarantee they stay available long-term. Content is semester-pinned (dates baked into the README/syllabus), so pulling this a year from now means filtering out stale scheduling info to find the parts that still apply.