metadata
dataset_info:
features:
- name: ids
dtype: string
- name: seqs
dtype: string
- name: texts
dtype: string
splits:
- name: train
num_bytes: 1009714080
num_examples: 1329150
download_size: 669995659
dataset_size: 1009714080
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
Heavily deduplicated version of Trembl (November 2023) with dense Natural Language protein descriptions.