metadata

dataset_info:
  features:
    - name: text
      dtype: string
    - name: id
      dtype: string
    - name: url
      dtype: string
  splits:
    - name: train
      num_bytes: 47250994990
      num_examples: 9672101
  download_size: 28314342419
  dataset_size: 47250994990
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
source_datasets: HuggingFaceFW/fineweb-edu
license: odc-by
task_categories:
  - text-generation
language:
  - en
tags:
  - 10BT
size_categories:
  - 1M<n<10M

fineweb-edu: 10BT sample

This the "10BT-sample" config of HuggingFaceFW/fineweb-edu with most of the redundant cols removed for efficiency reasons.

token counts

GPT-4 tiktoken token count:

        token_count
count  9.672101e+06
mean   1.001188e+03
std    1.834986e+03
min    3.800000e+01
25%    3.380000e+02
50%    6.090000e+02
75%    1.054000e+03
max    1.649670e+05

Total count: 9683.59 M tokens