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: 124335849835.91966
num_examples: 13377130
download_size: 42308647425
dataset_size: 124335849835.91966
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
A subset of FineWeb sample-100BT with sequence length >= 1024 when tokenized with the Llama 2 tokenizer (including special tokens)