a practical guide to fine-tuning open-source llms using peft and qlora — from dataset curation and formatting to training and quantised inference.

implementing ddpm in pytorch from first principles — the math behind forward and reverse diffusion, noise schedules, and training a small image model.