---
license: mit
language:
- en
library_name: graf
tags:
- graf
- paragraph-segmentation
- token-classification
- news
- encoder
- modernbert
- on-device
pipeline_tag: token-classification
base_model: jhu-clsp/ettin-encoder-17m
metrics:
- f1
---

# Graf

Paragraphs will be stacked until morale improves.

A small encoder that recovers paragraph breaks in flattened news text.

| | |
| --- | --- |
| **Task** | Paragraph break recovery (token classification) |
| **Size** | ~19 MB weights (trimmed vocab, int8 embeddings, fp16 rest) |
| **Runtime** | CPU/GPU (PyTorch); browser (ONNX Runtime WASM) |
| **Weights** | [`main`](https://huggingface.co/localangle/graf/tree/main) · [`web/`](https://huggingface.co/localangle/graf/tree/main/web) for WASM |
| **Demo** | [localangle/graf-demo](https://huggingface.co/spaces/localangle/graf-demo) (in-browser) |

Graf does **not** rewrite the article. It scores eligible token ends as
`BREAK` / `NO_BREAK`, then splits the flattened string at predicted breaks.

> Flattened: *City council approved a library renovation… Construction starts in March…*
> → Graf inserts breaks between the recovered paragraphs.

## Try it

- **Live demo:** [localangle/graf-demo](https://huggingface.co/spaces/localangle/graf-demo) — paste flattened news, paragraphize in the browser.
- **Local UI:** from the [graf](https://github.com/cjdd3b/graf) repo, `graf-web --web-package checkpoints/graf-web`.

## Install

```bash
pip install "graf @ git+https://github.com/cjdd3b/graf.git"
# or, from a checkout:
pip install -e .
```

Python ≥ 3.11. Needs `torch`, `transformers`, `safetensors`, `huggingface_hub`.

## Usage

```python
from graf.infer import GrafParagraphizer

graf = GrafParagraphizer("localangle/graf")  # thr from graf.json (0.41)
result = graf.paragraphize(
    "City council approved a library renovation after public comment. "
    "Paulsen and Chen voted against the measure. Construction starts in March."
)
print("\n\n".join(result.paragraphs))
```

Load a local package folder the same way:

```python
graf = GrafParagraphizer("checkpoints/graf")
```

## Files

| File | Contents |
| --- | --- |
| `graf.json` | Format, threshold, kept vocabulary, source metadata |
| `weights.safetensors` | Trimmed encoder + head (int8 token embeddings, fp16 elsewhere) |
| `encoder/` | Ettin/ModernBERT config for the trimmed vocabulary |
| `tokenizer/` | Full tokenizer (ids remapped at inference via `vocab_map`) |
| `web/` | Browser package: `model.onnx` (~17 MB int8), `vocab_map.bin`, `graf-web.json`, `tokenizer/` |

PyTorch weights expand to fp32 at load; only storage is compressed. The `web/`
ONNX graph is what the static demo runs in WASM (soft cap 2048 tokens).

## Accuracy

Held-out Extra structure v0.2 test set (publisher-disjoint), package threshold
0.41 (student before compression; packaging keeps break agreement ≈ Jaccard 0.98):

| Metric | Value |
| --- | --- |
| Boundary F1 | 0.871 |
| Precision / Recall | 0.845 / 0.900 |
| PR-AUC | 0.948 |
| Body–body precision / recall | 0.829 / 0.894 |
| Median \|Δ block count\| | 2.0 |

## How it works

Flattened article text is packed into overlapping 2048-token windows. Eligible
token ends (positions that can host a paragraph boundary) are scored for
`BREAK`. Probabilities are averaged across windows, thresholded, and used to
reconstruct paragraph blocks.

The published package is a compressed [Ettin](https://huggingface.co/jhu-clsp/ettin-encoder-17m)
student (~17M) distilled for on-device / in-browser use.

## Limitations

- English news-style prose; other domains are unmeasured.
- Training text / Extra labels are not redistributed here.
- Exact full-article reconstruction is intentionally hard; treat boundary F1 /
  body–body metrics as the primary quality signal.
- Browser int8 ONNX drifts slightly vs the PyTorch package; the web meta
  threshold is nudged for recall (see `web/graf-web.json`).

## Built on

- [Ettin encoder 17M](https://huggingface.co/jhu-clsp/ettin-encoder-17m) (MIT)
  by JHU CLSP. Fine-tuned and compressed for Graf; see
  `THIRD_PARTY_NOTICES.md`.

## License

MIT for the Graf code and this package's weights, subject to the Ettin MIT
notice above.

## Citation

```bibtex
@software{graf_2026,
  title  = {Graf: Paragraph break recovery for flattened news},
  author = {Local Angle},
  year   = {2026},
  url    = {https://huggingface.co/localangle/graf},
}
```