metadata
dataset_info:
features:
- name: input_id_x
sequence: int8
- name: input_id_y
sequence: int8
splits:
- name: train
num_bytes: 7582970656
num_examples: 17070828
download_size: 4615653058
dataset_size: 7582970656
Dataset Card for "tokenized-total-512-reduced"
This dataset contains truncated tokenized protein sequences and their corresponding 3Di structure as stated in the Foldseek paper. Redundancy reduction and data sequence filtering was performed by Dr. Michael Heinzinger and Prof. Dr. Martin Steinegger. The tokenizer used to encode the sequences can be found here