metadata
dataset_info:
features:
- name: input_ids
dtype:
array2_d:
shape:
- 256
- 4
dtype: int64
- name: position_ids
dtype:
array2_d:
shape:
- 256
- 4
dtype: int64
- name: loss
dtype:
array2_d:
shape:
- 256
- 4
dtype: float32
splits:
- name: train
num_bytes: 57402192
num_examples: 2437
download_size: 14549836
dataset_size: 57402192
configs:
- config_name: default
data_files:
- split: train
path: data/train-*