Sourced from biomap-research/metal_ion_binding and prepared for Hugging Face datasets usage.
Parquet files are stored under data/ using Hugging Face split naming conventions
(train-, validation-, test-*).
Preprocess mode: minimal.
No max sequence length filter was applied.
Validation handling: upstream did not provide a validation split, so validation was sampled from train using seed 1957723 and matched… See the full description on the dataset page:
https://huggingface.co/datasets/swhitfield/biomap-research-metal_ion_binding.
