metadata
dataset_info:
features:
- name: text
dtype: string
- name: loss_mask
sequence: int64
- name: input_len
dtype: int64
splits:
- name: train
num_bytes: 3191732424
num_examples: 900000
- name: test
num_bytes: 354656878
num_examples: 100000
download_size: 226528509
dataset_size: 3546389302
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
- split: test
path: data/test-*