Fern

hlb-CIFAR10. Train to 94% on CIFAR-10 in <6.3 seconds on a single A100. Or ~95.79% in ~110 seconds (or less!)

hlb-gpt. Minimalistic, extremely fast, and hackable researcher's toolbench for GPT models in 307 lines of code. Reaches <3.8 validation loss on wikitext-103 on a single A100 in <100 seconds. Scales to larger models with one parameter change (feature currently in alpha).

fileshare. A misc space to hold (smallish) files to share with others online

modded-nanogpt. NanoGPT (124M) in 5 minutes