metadata
dataset_info:
features:
- name: text
dtype: string
- name: id
dtype: string
- name: dump
dtype: string
- name: url
dtype: string
- name: date
dtype: string
- name: file_path
dtype: string
- name: language
dtype: string
- name: language_score
dtype: float64
- name: token_count
dtype: int64
splits:
- name: train
num_bytes: 686749697.4094799
num_examples: 200000
download_size: 420184700
dataset_size: 686749697.4094799
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
200k samples from Fineweb, around 140M tokens