ChrisHayduk/nanofold-public

NanoFold Public NanoFold Public is the public train/validation portion of the nanoFold protein-folding benchmark. It packages a compact, fixed, auditable subset of OpenProteinSet/OpenFold-derived protein structure training data for fast iteration on data-efficient folding models. The dataset has 10000 train chains and 1000 public validation chains. Each row is one protein chain. The original processed .npz tensors are unrolled into Hugging Face Dataset columns so users can load… See the full description on the dataset page: https://huggingface.co/datasets/ChrisHayduk/nanofold-public.