Deep Learning Basics with PyTorch — Study Notebooks

Study notes and code for Deep Learning Basics with PyTorch by Dr. Yves J. Hilpisch (The Python Quants GmbH). Chapter notebooks reimplement the material — foundations in NumPy, then PyTorch — with exercise sets per chapter and a California Housing capstone.

What's inside

• Ch 1–4 — foundations, deliberately NumPy-first
• Ch 5–8 — tensors, autograd, training loops, nn.Module
• Ch 9–12 — pipelines, regularization, CNNs, AMP, checkpointing
• Ch 13–15 — attention, transformers, large-model training
• exercises_challenges/ — exercise sets (Ch 1–15)
• capstone_california_housing/ — data → baselines → MLP → stretch goals

Capstone Notebooks — Run in Sequence

Notebooks 01–06 must be run in order. Notebooks 04–06 load fitted models and preprocessors produced by earlier notebooks (ridge_best.pkl, hgb_model.pkl, preprocessor.pkl). These artifacts are ignored by git and regenerated on each full-sequence run. If running individual capstone notebooks, first run 01 → 03 to generate the required artifacts.

All 36 notebooks are designed to execute on CPU. Validate locally with notebook_workflow.sh or jupyter notebook. Gradient-accumulation and checkpoint-resume are verified against uninterrupted baselines.

Setup

conda env create -f environment.yml
conda activate pytorch_dl_apps
jupyter lab

Attribution & License

Pedagogical credit for the book belongs to Dr. Yves J. Hilpisch. Code in this repository is MIT-licensed; the license does not cover the book's content or the official companion code.

Maintainer: Francisco Salazar