---
dataset_info:
  features:
  - name: input_ids
    sequence: int32
  splits:
  - name: train
    num_bytes: 5890822600
    num_examples: 1436786
  download_size: 2935508857
  dataset_size: 5890822600
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
---

Pre-tokenized dataset for Gemma-2 created from the first 2 data files of [HuggingFaceFW/fineweb](https://huggingface.co/datasets/HuggingFaceFW/fineweb) by SAELens.

This dataset has a 1024 context size and roughly 1.5B tokens.