README.md

---
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: token_count
      dtype: int64
    - name: dataset
      dtype: string
  splits:
    - name: train
      num_examples: 17781952
license: odc-by
language:
  - en
size_categories:
  - 10M<n<100M
tags:
  - pretraining
  - smol-data
pretty_name: FinePDFs-Edu 100BT
---

# FinePDFs-Edu 100BT

A ~100 billion token English subset of [FinePDFs-Edu](https://huggingface.co/datasets/HuggingFaceFW/finepdfs-edu) (eng_Latn split), created for efficient pretraining experiments.

Part of the [Smol-Data](https://huggingface.co/collections/HuggingFaceFW/smol-data) collection — tried and tested mixes for strong pretraining.

## Dataset Description

This dataset was created by randomly sampling from the English split of FinePDFs-Edu (~136B tokens) to produce a ~100B token subset. Since the source is only ~136B tokens and the target is 100B, the sampling rate is high (~77%), meaning most of the source data is retained.

A pre-shuffled version is available at [HuggingFaceFW/finepdfs_edu_100BT-shuffled](https://huggingface.co/datasets/HuggingFaceFW/finepdfs_edu_100BT-shuffled).

## How It Was Created

The dataset was generated using [datatrove](https://github.com/huggingface/datatrove) with the [smol_data.py](https://github.com/huggingface/datatrove/blob/main/examples/smol_data.py) script. The pipeline reads from the source dataset in streaming mode, applies a `SamplerFilter` to downsample, and writes the result back to the Hugging Face Hub.

## Usage

```python
from datasets import load_dataset

ds = load_dataset("HuggingFaceFW/finepdfs_edu_100BT", split="train", streaming=True)
for sample in ds:
    print(sample["text"][:200])
    break
```

## Citation

```bibtex
@misc{niklaus2026smoldata,
      title={SmolData},
      author={Joel Niklaus and Hynek Kydl{\'\i}{\v{c}}ek},
      year={2026},
      publisher={Hugging Face},
      journal={Hugging Face repository},
      howpublished={\url{https://huggingface.co/collections/HuggingFaceFW/smol-data}}
}
```