---
dataset_info:
  features:
  - name: input_ids
    sequence: int32
  splits:
  - name: train
    num_bytes: 5007987636.0
    num_examples: 152813
  download_size: 2195659342
  dataset_size: 5007987636.0
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
---

# MiniPile Tokenized for Llama 3 (~1.2b tokens)
This is a pre-tokenized version of the [JeanKaddour/minipile](https://huggingface.co/datasets/JeanKaddour/minipile) dataset using the Llama3 tokenizer.


## Licensing
Identical to original MiniPile dataset.


## Tokenization details
Tokenization was done using [SAELens](https://github.com/jbloomAus/SAELens) (3.11.0) with these settings:

- Context size: <em>8192</em>
- Shuffled: <em>yes</em>
- Begin batch token: <em>"bos"</em>
- Begin sequence token: <em>none</em>
- Sequence separator token: <em>"eos"</em>