Notebooks

Session Notebooks

Every technical session has a runnable companion notebook, generated from that session’s reading — the same code, shipped without stored outputs so that you run every cell yourself. Open in Colab runs it on the free tier (enough for all of them); view renders it on GitHub without running. They all live in the public repo dl-f2026-notebooks.

Notebooks are demos for the readings, not assignments — homework starters are zips attached to each assignment on Canvas. The guided review (S15), the alignment overview (S21), and the two code-defense days (S28–S29) ship no notebook by design.

Session	Topic	Run	View
S01	Why Deep Learning Works Now	Open in Colab	view
S02	From Linear Models to MLPs	Open in Colab	view
S03	Losses and Gradient Descent	Open in Colab	view
S04	Backpropagation I	Open in Colab	view
S05	Backpropagation II	Open in Colab	view
S06	Optimizers	Open in Colab	view
S07	Regularization, Initialization, Normalization	Open in Colab	view
S08	PyTorch in Practice	Open in Colab	view
S09	Convolution	Open in Colab	view
S10	From LeNet to ResNet	Open in Colab	view
S11	Transfer Learning and Data Augmentation	Open in Colab	view
S12	Training in Practice	Open in Colab	view
S13	RNNs and LSTMs	Open in Colab	view
S14	Attention	Open in Colab	view
S15	Guided Review of S01–S13	—	—
S16	Transformer I	Open in Colab	view
S17	Transformer II	Open in Colab	view
S18	LLM Pretraining	Open in Colab	view
S19	Fine-Tuning & LoRA	Open in Colab	view
S20	Decoding, Prompting, and In-Context Learning	Open in Colab	view
S21	Alignment: SFT, RLHF, DPO	—	—
S22	ViT & Multimodal	Open in Colab	view
S23	Autoencoders and VAEs	Open in Colab	view
S24	GANs	Open in Colab	view
S25	Diffusion Models	Open in Colab	view
S26	Efficiency and Scale	Open in Colab	view
S27	Evaluation and Safety	Open in Colab	view

If a notebook and its reading ever disagree, the reading is current and the notebook is about to be regenerated — report it and it will be fixed.