Wandering off the main path.

Training a small transformer to generate 16x16 pixel art characters by de-masking pixels step by step.

A simple MLP on MNIST to see how batch, layer, and group normalization differ in training speed, accuracy, and neuron activations.

Building a discrete diffusion transformer (DiT) from scratch and training it to generate Shakespeare-like text, one denoising jump at a time.

A follow-up to the 2-D diffusion notebook: steering generation with labels, from vanilla conditioning to classifier-free guidance.

A trimmed, annotated companion to the 2-D diffusion notebook — with an ODE-vs-SDE sampling comparison.