Hacker News Reader

is there a kind of nanogpt for diffusion language models? i would love to understand them better

This video has a live coding part which implements a masked diffusion generation process: https://www.youtube.com/watch?v=oot4O9wMohw