qleap/MPK1_dataset_by_NAGISA_V4
MPK1 dataset by NAGISA V4 Shogi teacher positions generated by attic-gensfen with the NAGISA V4 HalfKA-2304 evaluation function, carried in MPK1 — one record per searched position, before duplicate boards are folded into one. This is the input behind qleap/Knowledge_distilled_dataset_by_NAGISA_V4: no folding, no deduplication, no policy normalisation, no shuffling. 14,101,066 games, 1,372,612,150 positions. This is not the whole corpus its generator wrote.… See the full description on the dataset page: https://huggingface.co/datasets/qleap/MPK1_dataset_by_NAGISA_V4.
docs: drop the train/val split prescription
Cut games at ply 512 and drop the games the generator never finished (card)
Cut games at ply 512 and drop the games the generator never finished
Fold duplicate boards and repack (card)
Fold duplicate boards and repack
initial commit