Data

Tools

Indexes

Applications

Experiments

Open Source Science

GitHub / tysam-code 2 projects

Fern

Projects

Updated 6 months ago

hlb-gpt • Science 44%

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).

Updated 6 months ago

hlb-cifar10 • Science 44%

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