Chess Pre-to-Post — Pretraining Corpus v1 (20B)

Raw tokenized pretraining data for the Chess Pre-to-Post project, stored as sharded NumPy arrays (shard_XXXX/raw.NNNN.npy).

[!IMPORTANT] This is an earlier, smaller (20B) snapshot and is no longer maintained. The maintained version of this dataset is pavelslab-nyu/pretrain_v1_54B. Please use that version for any new work — it supersedes this one.

Maintained version

➡️ pavelslab-nyu/pretrain_v1_54B

Format

• Sharded raw token arrays: shard_0000/ … shard_0005/, each containing raw.NNNN.npy files of tokenized text.
• Load shards with numpy.load(...) and concatenate as needed for your data loader.

Citation

If you use this dataset, please cite the maintained version:

@misc{chess_pretrain_v1_54B,
  title        = {Chess Pre-to-Post Pretraining Corpus v1 (54B)},
  author       = {pavelslab-nyu},
  howpublished = {\url{https://huggingface.co/datasets/pavelslab-nyu/pretrain_v1_54B}},
  year         = {2026}
}