Study, reproduce, or adapt MiniMind small-language-model training stages, tokenizers, and checkpoints for bounded educational or research experiments.
