---
configs:
- config_name: bytelevel
  default: true
  data_files:
  - split: train
    path: bytelevel/*.parquet
- config_name: bytelevel-llm-data
  data_files:
  - split: ngram
    path: ngram/train-*
  - split: train
    path: bytelevel-llm-data/ngram/train-*
- config_name: bytelevel-subset
  data_files:
  - split: train
    path: bytelevel-subset/*.parquet
language:
- en
tags:
- language modeling
pretty_name: FineWebEDU 20B
size_categories:
- 10B<n<100B
dataset_info:
  config_name: bytelevel-llm-data
  features:
  - name: id
    dtype: string
  - name: input_ids
    sequence: int32
  - name: num_tokens
    dtype: int64
  - name: Entropy
    sequence: float64
  - name: Surprisal
    sequence: float64
  - name: Space Probability
    sequence: float64
  - name: EOS Probability
    sequence: float64
  splits:
  - name: train
    num_bytes: 226886
    num_examples: 2
  download_size: 160553
  dataset_size: 226886
---

# FineWebEDU 20B