pablovp8/

PockLigGPT-training-data

Preprocessed training assets for PockLigGPT.

chembl/ ├── train_chembl.bin └── val_chembl.bin

crossdocked/ ├── crossdocked_clean_pocket_selfies_with_tokens.parquet ├── per_residue_index.parquet └── per_residue_pack.npz.part-*

tokenizer/ └── meta_chembl_db_aa_2_proto4.pkl

The CrossDocked training Parquet contains pocket metadata, SELFIES, fixed-length token_ids, and offsets into the residue embedding stack. The split NPZ contains… See the full description on the dataset page: https://huggingface.co/datasets/pablovp8/PockLigGPT-training-data.