zhanshijinwat's tracked open-source repos, sorted by stars.
Train a 1B LLM with 1T tokens from scratch by personal