metadata
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
dataset_info:
features:
- name: input_ids
sequence: int32
- name: token_type_ids
sequence: int8
- name: attention_mask
sequence: int8
- name: labels
sequence: int64
- name: teacher_logits
sequence:
sequence: float64
- name: teacher_indices
sequence:
sequence: int64
- name: teacher_mask_indices
sequence: int64
splits:
- name: train
num_bytes: 156799366264
num_examples: 8620310
download_size: 0
dataset_size: 156799366264