---
dataset_info:
  features:
  - name: tokens
    dtype: uint16
    sequence_length: 2049
  splits:
  - name: train
    num_examples: 30873598
    num_shards: 2059
  - name: val
    num_examples: 34608
    num_shards: 3
license: odc-by
task_categories:
- text-generation
language:
- en
tags:
- pretokenized
- pre-packed
- fineweb-edu
- llama2
---

# prepacked-fineweb-edu-llama2-32K-T2048

Pre-tokenized and BOS-aligned best-fit packed version of [FineWeb-Edu](https://huggingface.co/datasets/HuggingFaceFW/fineweb-edu) for training with [looped nanochat](https://github.com/LoopedLMs/looped_nanochat).

Tokenized with the [Llama 2](https://huggingface.co/NousResearch/Llama-2-7b-hf) tokenizer (32,000 base vocab + 8 special tokens = 32,008).

## Stats

### Train split

| | |
|---|---|
| Source | [karpathy/fineweb-edu-100b-shuffle](https://huggingface.co/datasets/karpathy/fineweb-edu-100b-shuffle) (1,821 shards) |
| Total tokens | 63.26B |
| Total docs | 97.1M |
| Rows | 30,873,598 |
| Shards | 2,059 (`train-00000` to `train-02058`) |
| Rows per shard | ~15,000 |

### Val split

| | |
|---|---|
| Source | 2 shards (shard_01821–shard_01822) |
| Total tokens | 70.9M |
| Total docs | 106,496 |
| Rows | 34,608 |
| Shards | 3 (`val-00000` to `val-00002`) |

### Common

| | |
|---|---|
| Row length | 2,049 (T+1, seq_len=2048) |
| Format | Parquet, `uint16` token column |
| Tokenizer | Llama 2 32K ([NousResearch/Llama-2-7b-hf](https://huggingface.co/NousResearch/Llama-2-7b-hf)) |
| BOS token | `<s>` (id=1) |
| Packing | Best-fit, BOS-aligned, buffer_size=1000 |

Each row is a packed sequence of multiple BOS-delimited documents, cropped/filled to exactly 2,049 tokens. Rows are shuffled within each shard.