This repository provides a concise implementation of DDPM-style diffusion models for 2D datasets, designed as a learning resource rather than a production tool. It includes both the forward diffusion process (gradually adding noise) and the reverse denoising process (learning to recover data distributions). The README documents ablation studies on hyperparameters including learning rate, model size, number of timesteps, and variance schedules, with visualizations showing how each affects the quality of generated 2D point distributions.